The Effects of Firms’ Lobbying on

Resource Misallocation∗

Federico Huneeus† In Song Kim‡

Princeton MIT

First Draft: October 29, 2018This Draft: September 13, 2019

Abstract

We study the causal effect of firms’ lobbying activities on the misallocation of resources

through the distortion of firm size. To address the endogeneity between firms’ lobbying

expenditure and their size, we propose a new instrument. Specifically, we measure firms’

political connections based on the geographic proximity between their headquarter locations

and politicians’ districts in the U.S., and trace the value of these networks over time by

exploiting politicians’ assignment to congressional committees. We find that a 10 percent

increase in lobbying expenditure leads to a 3 percent gain in revenue. To investigate the

macroeconomic consequences of these effects, we develop a heterogeneous firm-level model

with endogenous lobbying. Using a novel dataset that we construct, we document new

stylized facts about lobbying behavior and use them, including the one from the instrument,

to estimate the model. Our counterfactual analysis shows that the return to firms’ lobbying

activities amounts to a 22 percent decrease in aggregate productivity in the U.S.

Keywords: Firm-level lobbying, misallocation, aggregate productivity, political connections

JEL Codes: D22, L11, L25, P16

∗We thank Charles Cameron, Allan Collard-Wexler, Emanuele Colonnelli, Remi Cura, Jason Davis, John de

Figueiredo, Jan De Loecker, Ted Enamorado, Alessandra Fenizia, Thomas Fujiwara, Gene Grossman, Matias Iaryc-

zower, Oleg Itskhoki, Gleason Judd, Dean Knox, Ilyana Kuziemko, Nolan McCarty, Atif Mian, Eduardo Morales,

Christopher Neilson, Ezra Oberfield, Mounu Prem, Steve Redding, Richard Rogerson, Esteban Rossi-Hansberg,

Juan Pablo Xandri as well as seminar participants at Duke University, Georgetown University, and Princeton Uni-

versity, for helpful discussions and comments. Eliza Riley provided superb research assistance. Huneeus is grateful

to financial support from the International Economics Section (IES), the Simpson Center for the Study of Macroe-

conomics, the Program for Latin American Studies (PLAS) and the Fellowship of Woodrow Wilson Scholars at

Princeton University. Kim acknowledges financial support from the National Science Foundation (SES-1725235).†Department of Economics, Princeton. Email: fhuneeus@princeton.edu, URL: http://www.fedehuneeus.com.‡Department of Political Science, Massachusetts Institute of Technology, Cambridge MA 02139. Phone: 617–

253–3138, Email: insong@mit.edu, URL: http://web.mit.edu/insong/www/.

1 Introduction

Distortions in the allocation of resources between firms can reduce aggregate productivity in

an economy (Hsieh and Klenow, 2009). Researchers have identified several channels through

which firms’ decision making influence this misallocation. For instance, by charging prices above

marginal costs, firms can produce less than efficiently (Baqaee and Farhi, 2017); by saving and thus

accumulating capital, firms can avoid financial constraints (Moll, 2014); and by choosing different

buyers, firms can influence the techniques other firms use to produce (Boehm and Oberfield, 2018).

Yet, an important dimension of firms’ decisions that is often overlooked in studying misallocation

is their capacity to influence policy-making directly through lobbying. Politically active firms may

obtain policy benefits at the expense of other firms (Khwaja and Mian, 2005; Kang, 2015), which,

in turn, could allow them to survive and grow more than otherwise.

In this paper, we study the causal effect of firms’ lobbying activities on the misallocation of

resources, through firms’ influence on policies that affect their size. We begin by documenting a

set of new facts about firms’ lobbying behavior in the U.S. In doing so, we construct a comprehen-

sive dataset of firm-level lobbying covering all lobbying activities in the U.S. from 1999 to 2018.

This dataset not only includes firms’ lobbying expenses but also identifies to which congressional

committees the lobbying activity is being exerted, by connecting it through detailed information

on all lobbied bills since the 106th Congress. We merge this dataset with standard economic

characteristics of public firms and show, among other facts, that firms that lobby more tend to

be bigger. This is consistent with the idea that lobbying leads to private benefits for politically

active firms.

To address the endogeneity in this relationship we build a new instrumental variable (IV).

We exploit exogenous variation in the value of firms’ connections with politicians by tracing the

assignment of those politicians to different congressional committees over time. This variation will

affect the returns to lobbying differentially as firms are heterogeneously exposed to committees

according to their own characteristics, such as which products they produce. The identification

assumption is that firms cannot influence committee membership. Thus, we follow the strategy of a

standard shift-share design (Adao et al., 2018), in which the share is the importance of a committee

for a firm, and the identification comes from the shift in committee membership of politicians who

1

are connected to those firms. We measure these connections based on the geographic proximity

between firms’ headquarter locations and politicians’ electoral districts. We find that a 10 percent

increase in lobbying expenditure leads to 3 percent revenue gains. Furthermore, our IV estimates

are an order of magnitude larger in absolute value than the OLS ones, highlighting the importance

of addressing the endogeneity in the relationship between lobbying expenses and firms’ revenues.

In order to understand the macroeconomic implications of this result, we develop a heteroge-

neous firm-level model with endogenous lobbying. The model features standard ingredients from

firm-level models such as heterogeneity in productivity, selection into production, and endogenous

entry. It also features endogenous lobbying activity. Firms self-select into lobbying by paying a

fixed cost, in the spirit of how selection works in Melitz (2003). Given this selection, firms choose

how much to lobby in order to gain policy benefits that provide revenue gains. The key mapping

in the model is between firms’ lobbying expenditures and these policy benefits. Although we im-

pose a functional form assumption for this mapping in our quantitative analysis, we provide one

micro foundation for this assumption through a simple game between a policymaker and firms

(Grossman and Helpman, 1994). The policymaker cares about the household’s utility as well

as lobbying expenditures. The latter may happen because the policymaker uses those resources

for legislative activities (Hall and Wayman, 1990; Hall and Deardorff, 2006). We find that the

policymaker rewards firms’ lobbying differentially as long as it is not possible for her to perfectly

substitute individual lobbying expenses. Hence, our model captures the empirical regularity that

politicians diversify the quid pro quo with their connections to firms in order to reduce the salience

and risks of firms’ influence.

We support key assumptions in the model by documenting several consistent facts that emerge

from the firm-level lobbying dataset. First, there is a strong selection into lobbying. Only around

12 percent of public firms lobby and these firms are significantly bigger than non-lobbyists. Second,

political activity is significantly persistent both in terms of lobbying expenses, and entry and exit

from lobbying. Third, lobbying behavior seems to be more consistent with firm-level lobbying

rather than sector-level. Specifically, business organizations account for but a small fraction of

total lobbying expenses while firms spend significantly larger amounts. We also find that firms

individually lobby on congressional bills that are concerned with very narrow policy issues that

directly affect them (e.g., a policy toward a specific product). In fact, the median number of special

2

interest groups that lobby on any given lobbied bill in the last 20 years is just two. The salience

of individual firm-level lobbying activity motivates our modeling assumption in Section 3 that

considers lobbying as a private activity of firms, rather than a coordinated effort that requires

collective actions among many firms. We also use the facts together with the moments from

firms’ size distribution, firms’ lobbying activity, and the aforementioned instrument to estimate

the model with a simulated method of moments.

Finally, starting from the estimated model, we perform a series of counterfactuals to understand

the macroeconomic consequences of firms’ lobbying activities. We show that firms’ lobbying

expenses reduce aggregate productivity by 22 percent relative to an economy where the return

to lobbying is set to zero. This reduction comes from two sources. The first is that reducing

lobbying implies a decline in the dispersion of firms’ marginal revenues of inputs, which improves

the allocation of resources. The second is that, through a general equilibrium effect, wages decline

so that entry becomes cheaper increasing the number of firms in the economy. This indirect effect

accounts for around 41 percent of the total effect, highlighting the importance of the model for

understanding the aggregate effects of firms’ lobbying activities.

We contribute to two strands of the literature. First, we connect to the literature of the mis-

allocation of resources between firms pioneered by Restuccia and Rogerson (2008) and Hsieh and

Klenow (2009). That literature has studied different margins of firms’ decision-making that influ-

ence the misallocation of resources such as pricing decisions in output markets (Baqaee and Farhi,

2017), financial frictions in capital markets (Midrigan and Xu, 2014) and contract enforcement

in intermediate input markets (Boehm and Oberfield, 2018), to name a few. Nevertheless, this

literature has missed an important dimension of firms’ decision-making, namely, their influence on

policy through lobbying activity. An exception to this is Arayavechkit et al. (2018), who look at

the effect of lobbying on capital misallocation. We contribute to this literature by implementing,

to the best of our knowledge, the first quantitative evaluation of the effect of firms’ lobbying activ-

ities on the misallocation of resources through the distortion of firm size. To do this, we develop

a general equilibrium firm model that features endogenous lobbying and estimate this model to

evaluate the macroeconomic effects of firms’ lobbying activities.

Next, we contribute to the political economy literature on corporate lobbying (Hansen and

Mitchell, 2000; Ansolabehere et al., 2002). Specifically, our study explains why firms get bigger as

3

a result of lobbying. This is in contrast to the conventional focus on the opposite causal direction

whereby researchers investigate how firms with different sizes tend to have different propensities

to engage in individual lobbying activities (Bombardini, 2008; Bombardini and Trebbi, 2012; Kim,

2017). To this literature, we make three contributions. First, we provide a new instrument to

identify the causal effect of firms’ political activity on their economic performance. The instrument

exploits exogenous variation to the value of firms’ political connections via changes in committee

memberships of politicians. The instrument uses an identification strategy that is similar to recent

research using committee assignments to obtain causal identification in studying firms’ political

activities (Bertrand et al., 2018; Fouirnaies and Hall, 2017; Powell and Grimmer, 2016). Second,

we quantify not only the firm-level effects of lobbying but also its macroeconomic effects. We find

significant private returns to lobbying (Richter et al., 2009; Kang, 2015) as well as a quantitative

evaluation of how politically connected firms may be responsible for inefficiencies in the U.S.

economy. Finally, we build a novel dataset that contributes to the rapidly growing empirical

literature that examines interest group lobbying (De Figueiredo and Richter, 2014). Our dataset

covers the universe of lobbying activities since 1999 and is matched to activities of other political

actors such as firms and politicians across various sectors and committees. We find that firm-level

lobbying expenditures are significantly larger than that by sectoral-organizations.

The remaining of the paper is organized as follows. The next section describes the data and

documents a set of novel stylized facts about firms’ lobbying behavior. Section 3 presents the

model. Section 4 implements the instrumental variable strategy, estimation of the model, and

counterfactual analysis. Section 5 concludes.

2 Data

We construct a novel database that connects firm-level economic activities to their political be-

havior for all publicly traded firms in the U.S. from 1999 to 2018. The Lobbying Disclosure Act

(LDA) of 1995 requires lobbyists to disclose their “lobbying activities”1 on behalf of their clients.2

We parse more than one million original filings available from the Senate Office of Public Records

1“Lobbying activities” are defined as “any oral or written communication (including an electronic communica-tion) to a covered executive branch official or a covered legislative branch official that is made.” The full list of thecovered federal agency names is available from the Office of the Clerk, U.S. House of Representatives.

2If a firm has its own in-house lobbying department, it should register and file lobbying reports indicating thatthey are “self” filing. In our sample, about 85% of lobbying is outsourced.

4

(SOPR). Each report contains information on the client firms, the identity of lobbyists, the to-

tal amount of lobbying expenditure in the corresponding period, list of issues lobbied, whether

lobbying activity was in-house or not, and lobbied legislative bills.3

Note that compliance of lobbying activity is closely monitored and enforced. Although the

contents of lobbying as well as the incurred expenses are based on a good faith description and

estimates by lobbyists, it is annually audited by the Government Accountability Office (GAO).

According to the 2014 audit report by GAO, 90% of lobbyists filed lobbying reports as required,

and 93% could provide documentation related to the expenses.4 Any lobbyist who fails to comply

with the legal requirements will be subject to $200,000 fine or up to 5 years of imprisonment, or

both as of 2015. Furthermore, lobbyists must immediately file an amendment of original filing

if they are notified of any defect or they omitted any relevant information on lobbying. Indeed,

lobbying information available from the reports has become a reliable source to study lobbying

in the literature (e.g., Ansolabehere et al., 2002; Bombardini and Trebbi, 2012; Bertrand et al.,

2014).

Our dataset is unique in two dimensions. First, we establish a direct link between lobbying

clients (i.e., firms) and the list of public firms. Indeed, the lack of standard company identifier

in the lobbying report has been a major constraint for conducting firm-level analysis of political

activities and their economic consequences. To the best of our knowledge, researchers have either

studied firms and trade associations at the level of sectors (up to 4-digit Standard Industrial

Classification) or focused primarily on a limited set of Fortune 500 and S&P 500 corporations

(e.g., Bombardini and Trebbi, 2012; Bertrand et al., 2018).5 We overcome this problem and study

political behavior of all publicly trading firms from 1999 to 2018. Specifically, we utilize natural

language processing, name entity matching algorithms, and manual matching to link 67,842 unique

lobbying client names to the list of public firm names and their standardized company identifiers

available from Compustat. Appendix A describes the details of this procedure. The lobbying

database as well as the firm identifiers (gvkey) is made publicly available through the webpage

3The LDA mandates lobbyists to disclose any congressional bill number, title, and the section of interest asso-ciated with lobbying.

4The 2014 GAO report on Lobbyist’s compliance with disclosure requirements is available fromhttp://www.gao.gov/products/GAO-15-310

5See Kim (2017) for an exception based on which we make further improvements disambiguating more firm

names covering the periods up to 2018.

5

(http://www.LobbyView.org).

Second, we measure the relative importance of congressional committees for each individual

firm i at year t by considering the complete list of bills that have been lobbied by the firm up to

t− 1. Specifically, we first identify the complete list of bills that have been lobbied by firm i. We

then check the committee c to which each bill is assigned and aggregate this information across

all lobbied bills. This procedure gives a measure of relative significance for each pair of firm and

committee across time, wict. Our approach differs from Bertrand et al. (2014) who assign relevant

issues6 to each congressional committees a priori. For example, the Senate Finance committee

is linked to the following lobbying issues: Unemployment, Trade, Taxation, Welfare, Retirement,

and Medicare/Medicaid.7 Note that multiple issues are mapped to multiple committees with equal

weights. They then consider the “issue overlap” between firms and politicians based on lobbied

issues and committee memberships. We improve upon this approach by exploiting the direct link

between bills that are actually lobbied by individual firms and committees where the bills are

considered. We also distinguish the relative importance of each committee for individual firms by

incorporating the frequency of bill-to-committee links. We provide further details of the measure

in Section 4.

2.1 Stylized Facts

In this section, we document six facts from the data that will guide the development of the model

in Section 3. We explore the relationship between firm characteristics and their lobbying activities.

Fact 1 Firm Lobbying is Relatively Rare. Lobbying is a relatively rare firm activity. Of the

7,646 public firms operating in the United States in 2017, only 766 firms engaged in lobbying. On

average, just 11.8 percent of public firms lobby across years. Table 1 illustrates the point more

broadly across two-digit NAICS sectors. We consistently find that lobbying is relatively rare. For

example, only about 5 percent of firms in Finance and Insurance sector (NAICS code 52) have

reported that they engaged in lobbying on any policy issues. Note that some firms are actively

6Section 15 of each report specifies the general issue areas of lobbying such as TAX (Taxation/InternalRevenue Code) and TRD (Trade (Domestic & Foreign). The full list of 79 issue codes is also available from theOffice of the Clerk, U.S. House of Representatives.

7For the complete list of mappings between congressional committees and issue codes used by Bertrand

et al. (2014), see https://assets.aeaweb.org/assets/production/articles-attachments/aer/app/10412/

20121147_app.pdf

6

NA

ICS

Cod

e#

Fir

ms

%L

ob

bie

d%

In-h

ou

seM

edia

nE

xp

ense

Exam

ple

Fir

m

Agr

icu

ltu

re,

For

estr

y,F

ish

ing/

Hu

nti

ng

11

26

20.4

7.6

$50,0

00

MO

NS

AN

TO

CO

Min

ing,

Qu

arry

ing,

and

Oil/G

asE

xtr

acti

on21

460

9.9

3.8

$40,0

00

RIO

TIN

TO

GR

OU

P(G

BR

)

Uti

liti

es22

289

22.7

15.3

$50,0

00

EN

EL

SP

A

Con

stru

ctio

n23

99

10.8

3.8

$30,0

00

FL

UO

RC

OR

P

Man

ufa

ctu

rin

g31-3

32,

930

15.8

6.5

$40,0

00

NE

ST

LE

SA

/A

G

Wh

oles

ale

Tra

de

42

220

8.1

3.4

$40,0

00

MC

KE

SS

ON

CO

RP

Ret

ail

Tra

de

44-4

5282

11.2

5.1

$60,0

00

CV

SH

EA

LT

HC

OR

P

Tra

nsp

orta

tion

and

War

ehou

sin

g48-4

9224

18.6

9.0

$45,0

00

EN

IS

PA

Info

rmat

ion

51

964

11.9

4.8

$50,0

00

AT

&T

INC

Fin

ance

and

Insu

ran

ce52

2,3

36

5.1

2.6

$50,0

00

UN

ITE

DH

EA

LT

HG

RO

UP

INC

Rea

lE

stat

ean

dR

enta

lan

dL

easi

ng

53

353

6.5

0.8

$40,0

00

BR

OO

KF

IEL

DA

SS

ET

MA

NA

GE

ME

NT

Pro

fess

ion

al,

Sci

enti

fic,

and

Tec

hn

ical

SV

C54

330

12.1

3.4

$40,0

00

AC

CE

NT

UR

EP

LC

Ad

min

/Was

teM

anag

emen

t/R

emed

iati

onS

VC

56

156

17.7

4.5

$40,0

00

MA

NP

OW

ER

GR

OU

P

Ed

uca

tion

alS

VC

61

35

24.6

8.3

$40,0

00

GR

AH

AM

HO

LD

ING

SC

O

Hea

lth

Car

ean

dS

oci

alA

ssis

tan

ce62

130

21.9

6.8

$50,0

00

HU

MA

NA

INC

Art

s,E

nte

rtai

nm

ent,

and

Rec

reat

ion

71

58

13.1

3.2

$30,0

00

LIV

EN

AT

ION

EN

TE

RT

AIN

ME

NT

Acc

omm

od

atio

nan

dF

ood

SV

C72

141

12.2

5.5

$50,0

00

SO

DE

XO

Oth

erS

ervic

es(e

xce

pt

Pu

bli

cA

dm

inis

trat

ion

)81

22

7.8

0.0

$40,0

00

SE

RV

ICE

CO

RP

INT

ER

NA

TIO

NA

L

Tab

le1:

Desc

ripti

ve

Sta

tist

ics

acr

oss

NA

ICS

2dig

itSect

ors

:T

his

table

pre

sents

des

crip

tive

stat

isti

csof

lobbyin

gac

tivit

ies

acro

ssal

lCompustat

firm

sin

21N

AIC

S2-

dig

itse

ctor

s.T

he

num

ber

sar

eth

eav

erag

esac

ross

year

sfr

om19

99to

2017

.F

orex

ample

,w

ese

eth

at,

onav

erag

e,th

ere

are

abou

t2,

930

firm

sin

am

anufa

cturi

ng

sect

or,

and

15.8

%of

them

are

found

tohav

elo

bbie

s,on

aver

age.

Am

ong

thes

e,ab

out

6.5%

has

in-h

ouse

lobbyin

gdep

artm

ent.

The

med

ian

lobbyin

gex

pen

ses

offirm

sin

each

sect

orra

nge

sfr

om$3

0,00

0to

$60,

000

per

year

.T

he

last

colu

mn

pre

sents

anex

ample

firm

who

lobbie

dfr

omea

chse

ctor

.

7

Tota

l Am

ount

(in

Bill

ion

Dol

lar)

02

46

8

Campaign ContributionLobbying Expenditure

1999

−200

0

2001

−200

2

2003

−200

4

2005

−200

6

2007

−200

8

2009

−201

0

2011

−201

2

2013

−201

4

2015

−201

6

Electoral Cycle

Figure 1: Money in Politics: Campaign Contributions vs. Lobbying Expenditures: Thisfigure compares the total amount reported to be spent for campaign contribution and lobbying. Itshows that, on average, lobbying expenditure is more than six times larger than campaign contribu-tion. We used data from FEC (available from http://classic.fec.gov/finance/disclosure/

ftpdet.shtml) to calculate the campaign contribution amount, which is the sum of “contributionor independent expenditure made by a PAC, party committee, candidate committee, or other fed-eral committee to a candidate during the two-year election cycle.” Note that we exclude individualcontributions to facilitate the comparison with the lobbying expenditure.

lobbying by establishing their own in-house lobbying department. However, this is also uncommon

as shown by the fourth column. For example, only 6.5% of manufacturing firms employ in-house

lobbyists.

Fact 2 More money is spent on lobbying than campaign contributions. Tullock (1972) asked

“why is there so little money in U.S. politics.” The so-called “Tullock’s Puzzle” is based on

the observation that campaign contributions in the 1970s sum to only about $200 million, which

is significantly smaller than the hundres of billions of dollars in public expenditures then. Re-

searchers still find that campaign contributions are relatively smaller than public spending of the

government (Ansolabehere et al., 2003). On the other hand, we find that lobbying expenditure

is significantly larger than campaign contributions. To be sure, money spent on lobbying is still

much smaller compared to the federal budget of about $4 trillion (as of 2016). However, as Fig-

ure 1 shows, we find that lobbying involves more money than campaign contributions made by

8

8 10 12 14 16

05

1015

2025

Log Lobbying Expenditure

Log

Sal

es

Coef = 0.984, P−value = 0.000, t = 28.13

Figure 2: Revenues and Lobbying Expenditure: This figure shows that firm’s size measuredby its sales is positively correlated with lobbying expense.

all PAC (political action committee), party committee, candidate committee, and other federal

committees combined. In Section 4, we will quantify the returns to lobbying.

Fact 3 Positive and robust correlation between firms’ revenues and lobbying activity. This

holds both in the extensive and intensive margin. As noted in Fact 1, firm-level lobbying is

relative rare. An important distinction that has been studied in the literature is that firms that

engage in lobbying tend to be larger than politically inactive firms in the extensive margin (Kim

and Osgood, 2018). Figure 2 shows that the positive correlation between firm size and lobbying

expenditure holds in the intensive margin as well. That is, conditional on lobbying, larger firms

tend to spend more money in lobbying.

Fact 4 Lobbying behavior is highly persistent. This holds both in the extensive and intensive

margin. Over time, lobbying activities are highly persistent. We examine this by tracking the

lobbying activities of all public firms that remain competitive in two consecutive years. The left

panel of Figure 3 shows that almost all firms that did not lobby in the previous year tend not to

lobby in the next year. On the other hand, firms that engaged in lobbying continue their political

activities. For example, more than 80% of firms that lobbied in 2016 continue lobbying in 2017.

9

0.5

0.6

0.7

0.8

0.9

1.0

Pro

port

ion

of F

irm

s P

ersi

sten

t in

Lobb

ying

Lobbying both t−1 and t

Not Lobbying both t−1 and t

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017

(a) Extensive Margin

8 10 12 14 16 18

810

1214

1618

Log Lobby (t − 1)

Log

Lobb

y (t

)

Coef = 0.922, P−value = 0.000, t = 157.88

(b) Intensive Margin

Figure 3: Persistence of Lobbying in the Extensive and Intensive Margins: We find thatfirm-level lobbying activities are persistent both at the extensive (whether lobbying) and intensivemargins (dollar amount conditional on lobbying). The left panel shows that firms that did notlobby in the previous year tends not to lobby in the following year as well (blue line). On theother hand, over 70% of firms tend to lobby in two consecutive years (red line). We note thatthere was a decrease in the intensive margin especially during the financial criss between 2007and 2009. The right panel shows that the intensive margin is highly correlated whereby lobbyingexpenses remains similar. The dotted line corresponds to the 45-degree line.

Note that this is a conservative measure of the persistence of lobbying as we focus exclusively on

two adjacent years. In fact, we observe a significant drop in the sticky behavior during the financial

crisis of 2007–2018, but the overall persistence becomes much higher as we allow for a wider window

over time. The right panel shows that there exists a positive and robust correlation between the

lobbying expenses in the internsive margin conditional on lobbying in both years. Moreover, we

find that the amounts of lobbying are also persistent in absolute values (indicated by the dotted

45 degree line). This is an important empirical fact that motivates our identification in Section 4

as we rely on the exogenous increases in the value of lobbying through political connections rather

than a strategic response in the amount of lobbying expenses at the firm-level when we evaluate

the economic effect of their lobbying activities.

Fact 5 Firm-level lobbying expenses are greater in amounts than the spending by sectoral

associations. To date, empirical studies of special interest group politics have focused primarily

10

Lobb

ying

Exp

endi

ture

(in

Bill

ion

US

D)

01

23

45

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

2014

2015

2016

2017

Public Firms (Compustat)Business Organization/Interest GroupsOthers (Including Private Firms)

Figure 4: Firm vs Sector Level Lobbying Expenditure: This figure compares the totallobbying expenses by firms and sectoral organizations. We first identify all public firms fromCompustat database (blue). To identify sectoral organizations (red), we included all lobbyingclients with NAICS code 813910 (“Business Associations”) along with others whose legal nameincludes “associations” or “ASSN.” All the other entities such as private firms and universities aregrouped as “Others” (green). We find that firm-level lobbying is significantly larger than that bysector-level lobbying.

on sector-level political activities such as campaign contributions (Goldberg and Maggi, 1999;

Gawande and Bandyopadhyay, 2000). Although lobbying through sectoral associations is highly

important, we find that firms’ individual lobbying activities are at least as prevalent as those by

sectoral organizations. In particular, Figure 4 shows that firm-level lobbying expenses (blue) are

in fact much larger than that by sectoral and business organizations combined (red). Note that

this excludes lobbying expenses by all private firms.

Fact 6 Most congressional bills are lobbied by only one or two interest groups. One of unique

contributions that we make is to identify lobbied congressional bills. According to the Lobbying

Disclosure Act of 1995, interest groups are legally required to report any congressional bills that

they have lobbied. For example, Bose Inc. reported that it lobbied on a Senate bill in the

109th Congress “A bill to reduce temporarily the duty on certain audio headphones achieving

11

All Bills

Number of Clients Lobbied0 100 200 300 400 500

010

000

2000

030

000

4000

050

000

6000

0

Tota

l Num

ber

of B

ills:

65,

047

Mean Number of Clients: 3.45Median Number of Clients: 2

Most Lobbied Bill: 113_HR1

Trade Bills

Number of Clients Lobbied0 20 40 60 80 100 120 140

020

040

060

080

010

0012

00

Tota

l Num

ber

of B

ills:

1,8

62

Mean Number of Clients: 3.21Median Number of Clients: 2

Most Lobbied Bill: 114_HR1314

Figure 5: Distribution of Number of Lobbyists per Bill (106th – 115th Congress): Thisfigure depicts the distribution of the number of lobbying clients that lobby on congressional bills.The left panel shows that the median number of clients that lobby on any given Senate or Housebill is two. The right panel shows a similar pattern for trade bills.

full-spectrum noise reduction” (S.2325). This is a bill that reflects highly specialized interest of

a particular firm, and in fact Bose Inc. was the only firm that reported to have lobbied on the

bill. Figure 5 shows that lobbying activities reflect narrow interest of political actors who tend

to lobby individually. Specifically, we find a highly skewed distribution of the number of interest

groups that lobby on any given bill with the overall median number equals to two. We find similar

patterns across various policy areas such as trade bills as shown by the right panel. Appendix C.1

shows the distribution across all 79 lobbying issues.8

3 A Theory of Firm-Level Lobbying

In this section, we develop a heterogeneous firm model with lobbying decisions. We proceed in

two steps in order to investigate misallocation of resources between firms through the creation

of distortions. The first, presented in Section 3.1, introduces a model that generalizes Hsieh and

Klenow (2009)’s framework along the lines of Melitz (2003). Specifically, we incorporate firm’s

decision to lobby, both whether to lobby or not (the extensive margin) and how much to spend on

lobbying activity (the intensive margin). In this first step, the mapping between lobbying effort and

8For each bill, we assign the most relevant lobbying issue by identifying the most frequent issue codes in the

lobbying report that a given bill appears.

12

distortions is taken to be given to facilitate our flexible exposition of the misallocation of resources

between firms. Section 3.2 presents the second step whereby we provide one micro foundation for

the mapping assumed in the previous step. We accomplish this goal by incorporating a simplified

version of Grossman and Helpman (1994)’s lobbying model.

3.1 Model with Exogenous Lobbying-to-Distortions Mapping

Overview of the Model This model is an extension of Hsieh and Klenow (2009)’s framework

along the lines of Melitz (2003). It is a multi-sector model that features standard ingredients

such as heterogeneous firms, selection into production and firm entry. An important key feature

that we introduce is that firms choose endogenously whether to lobby as well as how much they

spend on lobbying. Lobbying activity entails benefits and costs. The benefits are distortions that

exclusively help the firms’ revenues whereas the costs are the expenditures the firms have to spend

in order to lobby (which include a variable and a fixed component). The latter induces selection

into lobbying, and hence firms that lobby will obtain benefits through distortions at the cost of

aggregate misallocation. This is the main mechanism the model will explore.

Setup The economy is populated by a representative household and a mass X of firms. Each

firm produces a unique variety ω of a differentiated good. Firms are heterogeneous over id-

iosyncratic states in production, lobbying, and exogenous wedges. These states are denoted by

φ = (φP , φL, φτ ), where φP , φL and φτ is a Hicks-neutral productivity term, a lobbying produc-

tivity term, and an exogenous wedge term, respectively.9 Given the setup of the model, firms are

characterized by φ in the sense that all firms that produce varieties with the same φ behave in the

same way. There is an exogenous probability function over firm states denoted by F , with density

f . Similarly, there is an endogenous probability function over firm states, given by firm selection,

denoted by Fs, with density fs.10

Household The household supplies inelastically N units of labor while receiving firms’ profits

and revenues created by government policies. It has nested preferences first over different sectors

and over firms’ differentiated varieties within sectors:

9A quality demand shifter is ignored because it is standard to show that in this class of models, and with the

available data, one could identify separately the demand shifter from the Hicks-neutral productivity term.10Note that while the exogenous probability is the same across sectors, the endogenous probability function over

firm states varies across sectors because selection into sectors is heterogeneous across sectors.

13

Y =S∏s=1

Y θss , with

S∑s=1

θs = 1 (1)

Ys =

[∫(cs(φ))

σCs −1

σCs dFs(φ)

] σCsσCs −1

, (2)

where S is the number of sectors, θsSs=1 are the Cobb-Douglas shares. Ys is the aggregate demand

of sector s with a constant elasticity of substitution (CES), cs(ω), across the varieties within sector

s, and Ms is the endogenous mass of firms in sector s. Each sector has a different elasticity of

substitution, σCs > 1. Given this setup, households maximize their utility subject to their budget

constraint.

Technology Each firm produces output of a differentiated variety by combining variable inputs

and capital according to a Cobb-Douglas constant returns to scale (CRS) production function:

ys(φ) = φPs ns(φ)αNs ks(φ)α

Ks xs(φ)α

Xs ,

where ys(φ), ns(φ), ks(φ) and xs(φ) are output, employment, capital stock, and intermediate

inputs of firm φ in sector s, respectively, and αis are the Cobb-Douglas weights in sector s, where

CRS implies that αNs + αKs + αXs = 1.11 In order to produce, firms in each sector s have to spend

fPs units of labor.

Market Structure The market structure of this economy is monopolistic competition. This

is a standard assumption in the literature that implies that firms charge a constant markup

over marginal costs. Note that, we allow for heterogeneous markups across sectors, because the

elasticity of substitution in demand, cs(ω), is modeled to be different across sectors.

Distortions Firms face output distortions τs(·).12 These distortions imply departures from

optimal allocations, given the market structure of monopolistic competition. We assume these

distortions are collected as revenues by the government and rebated back to the household via a

11The model can be easily extended to include demand shifters and decreasing returns to scale in production.

This might be important given the caveats of Hsieh and Klenow (2009)’s framework described in Haltiwanger et al.

(2018). These extensions will be explored in future versions of this paper.12Similar objects are named wedges in the framework of Hsieh and Klenow (2009).

14

lump-sum transfer, T , thus keeping a balanced budget. These distortions can come from regulation

such as sales taxes. For the purpose of this paper, we need not take a stand on which are the

specific sources for these wedges. These wedges are defined by:

1 + τs(φ) =(φLs ls(φ)

)δs+ φτs , (3)

where ls(φ) are the resources allocated towards lobbying activity (which could be zero), δs is

a parameter that governs the curvature of the distortions-to-lobbying effort, and φτs is the exoge-

nous component of the distortions. Thus, there are two sources of distortions in this economy: An

endogenous one that comes from lobbying activity and an exogenous one. We include this compo-

nent to account for other possible sources of misallocation and thus not load all the misallocation

in the economy to lobbying activity.

Lobbying Decision Firms can decide whether to spend resources in lobbying activity. In order

to lobby, a firm φ in sector s has to spend fLs units of labor as fixed lobbying cost. This governs

the extensive margin of lobbying activity. Conditional on lobbying, φ has to choose how much

to spend on lobbying activity, ls(φ). When making this decision, it compares the benefits from

lobbying, which are given by the extra revenue provided by the distortion, and the variable cost

of spending ls(φ) resources on lobbying.

Market Clearing Conditions Market clearing conditions in this economy are characterized

by firms’ output, labor, intermediate inputs, and a government balanced budget constraint:

ys(φ) ≥ cs(φ) + xSs (φ), ∀s, ∀φ

N ≥S∑s=1

(ME

s fEs +

∫ (ns(φ) + fPs + 1

L(φ)(ls(φ) + fLs

))dFs(φ)

)∫xss(φ)dFs(φ) ≥

∫xs(φ)dFs(φ)

T ≡S∑s=1

∫τs(φ)rs(φ)dFs(φ),

where 1L(φ) is an indicator function equals to one if firm φ chooses to lobby, mSs (φ) is total output

sold to other firms, and rs(φ) is firm φ’s revenue. Note that we assume that the capital market

takes the interest rate as given (an assumption as if the US was a price taker in capital markets)

15

and also that the intermediate input market clears, industry by industry. This assumption is not

very important since we evaluate misallocation only within industries. We assume this because it

makes the model easier to solve.

Zero-Profit Conditions Given the fixed production and lobbying cost, firms’ production and

lobbying extensive margin decisions are characterized by the following zero-profit conditions (ZPC):

(ZPC-Production) πNLs (φ∗s) = 0 (4)

(ZPC-Lobbying) πLs (φ∗∗s ) = πNLs (φ∗∗s ), (5)

where πNLs (·) and πLs (·) are the profit functions if the firm does not lobby and does lobby,

respectively. Equation (4) says that if a firm has φ = φ∗s, then it will not lobby and produce zero

net profits from producing. Thus, since πNLs (·) and πLs (·) are increasing functions in its argument,

firms with φ < φ∗s do not find it profitable to produce. Conversely, firms with φ > φ∗s do find it

profitable to produce, but maybe not to lobby. Similarly, firms with φ = φ∗∗s choose to produce

and lobby, but gain zero net profits. Firms with φ < φ∗∗s choose not lobby whereas those with

φ ≥ φ∗∗s choose to lobby. Furthermore, firms with φ∗s ≤ φ < φ∗∗s choose to produce but not to

lobby. Thus, these ZPCs imply cutoffs in firms’ states that characterize firms’ extensive margin

decision into production and lobbying activity.

Free Entry Condition There is free entry (FE) in this model. Firms have to pay an entry cost

fEs in terms of labor, in order to have the option to take a draw of their state φ. The free entry

condition is characterized by the following:

(FE) V Es = 0,

where V Es = E

[Vs − fEs

]and Vs are the expected net and gross value of entry in sector s,

respectively.13 More details about the full solution of the model can be found in Appendix E.1.

Lobbying and Revenues Given the setup of the model, Proposition 1 summarizes the rela-

tionship between lobbying expenditures and revenues.

13Vs =∑∞

t=0(1− ηs)tπs,t, where ηs is the exogenous death rate of firms and πs,t is the average profit of firms in

sector s at time t.

16

Proposition 1. Using the first order conditions, the relationship between lobbying and revenues

is the following:

log rs(φ) = γ0 + (1− δs) log ls(φ)− δs log φLs (6)

Proof. All proofs are in Appendix E.3.

The result comes from the first-order condition of firms’ intensive margin decision on lobbying.

It says that the relationship between lobbying activity and revenues is log linear, with a return

1 − δs. The residual of this relationship is firms’ lobbying productivity φLs . Importantly, this

proposition shows why running a simple ordinary least squares (OLS) between these two charac-

teristics would induce a biased estimator of 1− δs, since corr(log ls(φ), log φLs ) 6= 0. The value of

this correlation would inform about the direction of the bias. One conjecture is that firms that are

more productive in producing are also more productive in lobbying. Under this conjecture, the

OLS estimate would underestimate the true effect of lobbying on revenues. We revisit this issue

in Section 4, but highlight for now that the model provides a clear interpretation of the positive

relationship between size and lobbying, while revealing the limitations of inference using naive

correlations between these two characteristics.

Lobbying and Misallocation Given the relationship between lobbying, distortions, and firm

outcomes, we now show how this influences aggregate productivity. Proposition 2 directly char-

acterizes the connection, extending the aggregation result from Hsieh and Klenow (2009).

Proposition 2. Aggregate output and sectoral productivity in this economy is given by the follow-

ing:

Y =S∏s=1

(ΦPs N

αNss KαKs

s XαXss

)θs(7)

ΦPs = M

1

σCs −1s︸ ︷︷ ︸Entry

(NPs

Ns

)αNs︸ ︷︷ ︸Fixed Costs

[∫ (φPs

TFPRs

TFPRs(φ)

)σCs −1

fs(dφ)

] 1

σCs −1

︸ ︷︷ ︸Aggregation of F irms′ Productivity

, (8)

where ΦPs is aggregate productivity in sector s, LPs is the total labor used directly in production

as opposed to paying for fixed costs, fs(·) is the equilibrium density of firms that produce in the

economy, TFPRs(φ) = pys(φ)φPs is revenue-productivity of firm φ in sector s, i.e. the market value

17

of firms’ productivities, and TFPRs is the average revenue-productivity across firms within sector

s. Expression (8) shows that aggregate productivity in this economy is influenced by three forces.

The first involves entry, the second the use of fixed costs in the economy and the third, how firms’

productivity and quality are aggregated. It is in this last term that one can see the influence of

distortions on aggregate productivity through distorting how much each firm is weighted in this

aggregation.14 Intuitively, in the absence of distortions, TFPRs(φ) = TFPRs and thus firms are

aggregated according to the weights given by the equilibrium density of firms, fs(·). In the presence

of distortions, this is no longer the case. Firms that have a higher output distortion, τs(φ), say

because they lobby more, will have lower marginal revenue products, and thus a lower revenue-

productivity, TFPRs(φ), than the average one from the sector they belong to. This implies that

those firms’ productivity will influence aggregate productivity more than what they should in the

absence of distortions. This is the mechanism we explore quantitatively in Section 4.2. Before

doing that, we show one way to micro found the assumption made in Equation (3) with respect

to how firms’ lobbying influences distortions and their revenues.

3.2 A Microfoundation of Lobbying-to-Distortions Mapping

Overview of the Model In the previous section, we took an exogenous mapping between firms’

lobbying effort and distortions. This section propose one micro foundation for this mapping based

on a game between the government and firms. The government cares about the household’s utility,

and thus about efficiency and firms’ lobbying expenditures. In exchange for lobbying expenditures,

the government is willing to give away efficiency by creating distortions. These distortions act as

private benefits for firms, for which firms are willing to incur lobbying expenses. By endogeneizing

the mapping between distortions and lobbying, this model proposes one micro foundation for the

misallocation of resources between firms. By giving more benefits to firms that lobby more, the

government introduces dispersion in the marginal revenue products of factors that firms spend on,

and thus on revenue total factor productivity, TFPRs(φ). Dispersion in this measure across firms

within sectors represents misallocation in this economy.

14This is only a partial equilibrium analysis because changes in the distortions might also affect how many

resources are used in fixed costs and how many firms enter. The general equilibrium effects of changes in lobbying

and distortions are postponed until the quantitative analysis in Section 4.2.

18

Setup The game between the government and firms consists of three stages. In the first, firms

choose whether to enter, whether to lobby, and how much to lobby. In the second stage, the

government chooses distortions given firms’ entrance and lobbying efforts. In the final stage,

firms choose how much to produce given the policies and the household chooses consumption.

The final stage can be thought as a regular firm model with distortions such as the one from

Hsieh and Klenow (2009). The difference here is that those distortions are endogenous to firms’

political activities, in a game between firms and the government. Given perfect foresight and no

uncertainty, we solve the model with backward induction.

Stage three of this game is the regular firm model with the structure already described in the

previous section.15 The only difference is that in stage three, there is no longer a lobbying decision.

By this stage, lobbying decision has already been made and distortions are already defined. Note

that distortions are exogenous at this stage. In stage two, the government solves the following

problem:

W = maxτs(·)

V C (py(φ), τ(φ)) + a

[∫ (φLl(φ)

)σL−1

σL f(dφ)

] σL

σL−1

︸ ︷︷ ︸L

(9)

s.t.

V C (py(φ), τ(φ)) =I − TP

(10)

∂y(φ)

∂τ(φ)= σC

y(φ)

1 + τ(φ)(11)

∂l(φ)

∂τ(φ)=

∂π(φ)

∂τ(φ), (12)

where V C (py(φ), τ(φ)) is the household’s indirect utility, L is a CES aggregator of lob-

byists’ expenditures and a is the relative weight that the government allocates to lobbyists in

comparison with the representative household. Government welfare is the sum between house-

hold’s welfare and the welfare it obtains from lobbying activity. Government can directly care

about lobbying activity for several reasons. The simplest one is that the lobbyists can provide

another source of income for the government. For the purpose of our analysis, we do not take

a stand on the source of this interest. We claim that an objective function like this can provide

15In order to simplify the exposition and stress more the intuition of this model, we assume one sector and

one factor of production (e.g., labor). Extending the model to a multi-sector and multi-factor environment is

straightforward.

19

one analytical micro foundation for the relevant mapping between lobbying effort and wedges.16

Equations (10) and (11) come from the household and firms’ problem in stage 3. Equation (12) is

a condition that says that firms are truth-telling in terms of how much they are willing to spend

on lobbying the government in return for an extra revenue of wedges. Note that this condition

is effectively using the optimality in the decision to lobby in the first stage of the game. This

condition is important because it avoids coordination issues that could arise otherwise, which are

beyond the scope of this paper.

Finally, in the first stage, firms choose whether to lobby and how much to spend conditional

on lobbying, and whether to enter the market.

Proposition 3. The solution to the problem stated in Equations (9)-(12) is the following:

τ(φ)

1 + τ(φ)= 1 + σC + a

φL

σC − 1

(φLl(φ)

L

) 1

σL(

1− F (φ∗∗)

1− F (φ∗)

)(13)

Proposition 3 provides three predictions of how the government allocates distortions in this

game. First, if the government does not value firms’ lobbying expenditures (a = 0), then τ(φ)/(1+

τ(φ)) = 1 + σC . That is, the government would still allocate a flat distortion within sectors to

fix the distortion created by constant markups. Second, if the government does value lobbying

(a > 0), then distortions are heterogeneous depending on how much lobbying activities firms

actually engage in. How much they vary across firms depends crucially on σL, the elasticity

of substitution of lobbying contributions. The higher σL, the easier the government substitutes

lobbying expenditures between firms, and thus the more concentrated lobbying expenditure is

across firms. In other words, the higher σL, the less τ(φ) varies with l(φ). In the limit, when

lobbying expenditures are perfect substitutes (σL → ∞), τ(φ) is independent from l(φ).17 The

intuition of these results are important. Why would the government have love for variety of

lobbying expenditure? One reason could be that lobbying entails political risks. Being subject to

the influence of only one lobbyist could be politically costly for the government because it would be

at the cost of the household’s welfare and it could be easier to identify that concentrated influence

16This welfare function is a generalization of the one used in Grossman and Helpman (1994). In fact, in the

limit σL → 1, for all sectors, it becomes the same welfare function where the government aggregates lobbying effort

linearly. Thus, our specification for the welfare function nests Grossman and Helpman (1994)’s.17Note that in this case, one would arrive to the specification of the government’s welfare in Grossman and

Helpman (1994).

20

and make opposition to it. Whereas, if the influence is dispersed across actors, it could be less

difficult for the household to form opposition to it because the interests would be spread. Thus,

the love for variety could arise due to a preference of the government to reduce political risks of

opposition to lobbying expenditure. This is how the model justifies heterogeneous distortions and

lobbying expenditures at the firm level. The facts shown in Section 2 are consistent with this view

of lobbying behavior. Finally, τ(φ)/(1 + τ(φ)) increases with the mass of firms lobbying, 1−F (φ∗∗)1−F (φ∗)

.

4 Empirical and Quantitative Analysis

This section describes the main empirical exercises implemented in the paper. It proceeds in three

steps. First, it describes the instrumental variable (IV) approach designed to causally address

how lobbying influences firm size. Second, it takes moments from the data, and estimates the

parameters of the model. Finally, it concludes with counterfactual analysis to investigate how

lobbying affects the misallocation of resources and aggregate productivity.

4.1 Causal Evidence of Lobbying Expenditure on Revenues

Lobbying Instrument The relationship between lobbying expenditures and firm size is subject

to a standard endogeneity challenge, which is shown explicitly in Proposition 1. As in a standard

productivity estimation, the effect of lobbying expenditure on revenues needs to control for the

productivity of lobbying. Since lobbying is chosen as a function of its productivity, for identifi-

cation one needs variation in lobbying behavior that is exogenous to variation in firms’ lobbying

productivity. In order to address this, we propose a new instrument that shifts the profitability

of lobbying, holding constant firms’ primitives. The instrument measures changes in the value of

firms’ political connections by exploiting (a) changes in politicians’ committee membership in the

U.S. Congress, (b) heterogeneity in firms’ exposure to committee activity, and (c) firms’ political

connections. It follows a standard shift-share design. The shifts come from politicians’ changes in

committee membership in the U.S. Congress, that affects firms heterogeneously because they are

subject to connections to different politicians and also different exposures to committees’ activities.

Formally, the instrument is defined as follows:

zit =∑

j∈Ωi

∑cwict−k︸ ︷︷ ︸Share

djct︸︷︷︸Shift

(14)

21

where i and t denote firms and years, Ωi is the set of politicians in firm i’s networks, wict−k is

the weight that firm i gives to committee c in period t − k, and djct is a dummy variable equal

to one if politician j is assigned to committee c in period t. Thus, the instrument exploits three

ingredients and its interactions: Ωi, wict−k and djct. We describe each in turn.

First, firm i’s political connections, Ωi, is defined by the co-location of i’s headquarters and

the politicians of that district. Politicians that represent the state where i’s headquarters is

located, belong to i’s connections. Second, committee weights, wict−k, represent how important a

committee is for a firm in terms of the frequency of committee assignments of the bills that are

directly lobbied by the firm. Formally, the weights are defined as follows:

wict−k =bict−k∑h biht−k

where bict−k is the number of bills that i lobbied and were assigned to committee c in year

t−k. Thus, wict−k measures the share of bills that firm i lobbied that are under the jurisdiction of

committee c relative to all the bills lobbied by i considered in all committees. In order to measure

this, we searched the entire lobbying reports to identify the bills that have been lobbied by each

individual firm, let alone the respective committee that each bill was assigned to.

Finally, djct measures the shift of how politicians move between committees.18 This shift

provides the identification of the instrument.19 The key identifying assumption is that changes of

politicians between committees is exogenous to firms characteristics and influence. Thus, a key

component of the instrument relies on politicians actually changing committees over time. The

churning of committee membership is presented in Figure 6. It describes how often politicians move

to a new committee (red) and how likely that they stay in the committees between two consecutive

congresses. The figure shows that changes of committees is frequent for both democrats and

republican senators. Quantitatively, the average probability of a politician changing a committee

between Congress is around 30 percent. This number is relatively constant across Congresses, as

Figure C.2 shows in the appendix.

To construct the instrument for firm i at time t, we rely on the following three simultaneous

variations. First, a politician needs to change committees over time. Second, that politician needs

18The committee assignment data is from Stewart and Woon (2011).19As opposed to the share providing the identification, as in the shift-share design in Goldsmith-Pinkham et al.

(2018).

22

Akaka, Daniel K. (HI)Baldwin, Tammy (WI)

Baucus, Max (MT)Bayh, Evan (IN)

Begich, Mark (AK)

Bennet, Michael (CO)

Biden, Joseph R., Jr. (DE)Bingaman, Jeff (NM)

Blumenthal, Richard (CT)Booker, Cory (NJ)

Boxer, Barbara (CA)Breaux, John B. (LA)Brown, Sherrod (OH)

Bryan, Richard H. (NV)Burris, Roland (IL)

Byrd, Robert C. (WV)Cantwell, Maria (WA)

Cardin, Benjamin L. (MD)

Carnahan, Jean (MO)

Carper, Thomas Richard (DE)

Casey, Robert P., Jr. (PA)Cleland, Max (GA)

Clinton, Hillary Rodham (NY)Conrad, Kent (ND)

Coons, Christopher (DE)

Cortez Masto, Catherine (NV)

Corzine, Jon Stevens (NJ)

Cowan, William (Mo) (MA)

Daschle, Thomas A. (SD)Dayton, Mark (MN)

Dodd, Christopher, J (CT)Donnelly, Joe (IN)

Dorgan, Byron (ND)

Duckworth, Tammy (IL)

Durbin, Richard J. (IL)Edwards, John (NC)

Feingold, Russel D. (WI)

Feinstein, Dianne (CA)Franken, Al (MN)

Gillibrand, Kirsten E. (NY)Graham, Bob (FL)Hagan, Kay (NC)Harkin, Tom (IA)

Harris, Kamala (CA)

Hassan, Maggie (NH)

Heinrich, Martin (NM)Heitkamp, Heidi (ND)

Hirono, Mazie (HI)

Hollings, Ernest F. (SC)

Inouye, Daniel K. (HI)

Jeffords, James M. (VT)Johnson, Tim (SD)

Kaine, Timothy (VA)Kaufman, Ted (DE)

Kennedy, Edward M. (MA)

Kerrey, J. Robert (NE)

Kerry, John F. (MA)

Kirk, Paul (MA)

Klobuchar, Amy (MN)Kohl, Herbert H. (WI)

Landrieu, Mary L. (LA)

Lautenberg, Frank R. (NJ)Leahy, Patrick J. (VT)

Levin, Carl (MI)

Lieberman, Joseph I. (CT)

Lincoln, Blanche Lambert (AR)

Manchin, Joe (WV)

Markey, Edward J. (Ed) (MA)

McCaskill, Claire (MO)

Menendez, Robert (NJ)

Merkley, Jeff (OR)

Mikulski, Barbara A. (MD)

Miller, Zell Bryan (GA)

Moynihan, Daniel P. (NY)

Murphy, Christopher (CT)Murray, Patty (WA)

Nelson, Clarence William (Bill) (FL)

Nelson, Earl Benjamin (Ben) (NE)Obama, Barack (IL)

Peters, Gary (MI)Pryor, Mark (AR)Reed, Jack (RI)

Reid, Harry (NV)

Robb, Charles S. (VA)

Rockefeller, John D., IV (WV)

Salazar, Kenneth Lee (CO)

Sarbanes, Paul S. (MD)Schatz, Brian E. (HI)

Schumer, Charles Ellis (Chuck) (NY)

Shaheen, Jeanne (NH)Specter, Arlen (PA)

Stabenow, Deborah Ann (MI)Tester, Jon (MT)

Torricelli, Robert G. (NJ)

Udall, Mark (CO)Udall, Tom (NM)

Van Hollen, Chris (MD)Walsh, John E. (MT)

Warner, Mark (VA)

Warren, Elizabeth (MA)Webb, Jim (VA)

Wellstone, Paul David (MN)

Whitehouse, Sheldon (RI)Wyden, Ron (OR)

106 107 108 109 110 111 112 113 114 115

Congress Number

Sen

ator

s (D

emoc

rat)

Abraham, Spencer (MI)

Alexander, Lamar (TN)Allard, Wayne (CO)Allen, George (VA)

Ashcroft, John (MO)Ayotte, Kelly (NH)

Barrasso, John A. (WY)

Bennet, Robert F. (UT)Blunt, Roy (MO)

Bond, Christopher S. (MO)

Boozman, John (AR)Brown, Scott (M

A)

Brownback, Samuel Dale (KS)

Bunning, James Paul David (KY)Burns, Conrad (MT)

Burr, Richard (NC)

Campbell, Ben Nighthorse (CO)

Capito, Shelley M. (WV)Cassidy, Bill (LA)

Chafee, John H. (RI)

Chafee, Lincoln Davenport (RI)

Chambliss, Saxby (GA)Chiesa, Jeffrey (NJ)

Coats, Dan (IN)

Coburn, Thomas Allen (OK)

Cochran, Thad (MS)

Coleman, Norm (MN)Collins, Susan (ME)

Corker, Bob (TN)Cornyn, John (TX)Cotton, Tom (AR)

Coverdell, Paul (GA)Craig, Larry E. (ID)

Crapo, Michael Dean (ID)Cruz, Ted (TX)

Daines, Steve (MT)

DeMint, James W. (SC)

DeWine, Michael (OH)

Dole, Elizabeth Hanford (NC)

Domenici, Pete V. (NM)

Ensign, John Eric (NV)

Enzi, Michael B. (WY)Ernst, Joni (IA)

Fischer, Deb (NE)

Fitzgerald, Peter G (IL)Flake, Jeff (AZ)

Frist, William H. (TN)Gardner, Cory (CO)Gorton, Slade (WA)

Graham, Lindsey O. (SC)Gramm, Phil (TX)Grams, Rod (MN)

Grassley, Charles E. (IA)Gregg, Judd (NH)

Hagel, Chuck (NE)Hatch, Orrin G. (UT)

Heller, Dean (NV)Helms, Jesse (NC)Hoeven, John (ND)

Hutchinson, Tim (AR)

Hutchison, Kay Bailey (TX)

Inhofe, James M. (OK)

Isakson, Johnny (GA)

Jeffords, James M. (VT)Johanns, Mike (NE)Johnson, Ron (WI)

Kennedy, John N. (LA)Kirk, Mark (IL)Kyl, Jon (AZ)

Lankford, James (OK)Lee, Mike (UT)

LeMieux, George (FL)Lott, Trent (MS)

Lugar, Richard G. (IN)Mack, Connie (FL)

Martinez, Melquiades R. (Mel) (FL)McCain, John (AZ)

McConnell, Mitch (KY)Moran, Jerry (KS)

Murkowski, Frank H. (AK)

Murkowski, Lisa (AK)Nickles, Don (OK)

Paul, Rand (KY)Perdue, David (GA)Portman, Rob (OH)

Risch, James (ID)Roberts, Pat (KS)

Roth, William V., Jr. (DE)Rounds, Mike (SD)Rubio, Marco (FL)

Santorum, Rick (PA)Sasse, Ben (NE)

Scott, Tim (SC)Sessions, Jeff (AL)

Shelby, Richard C. (AL)

Smith, Gordon H. (OR)

Smith, Robert C. (NH)

Snowe, Olympia J. (ME)Specter, Arlen (PA)Stevens, Ted (AK)

Strange, Luther (AL)

Sullivan, Daniel (AK)

Sununu, John E. (NH)

Talent, James Matthes (MO)

Thomas, Craig (WY)

Thompson, Fred (TN)Thune, John (SD)

Thurmond, Strom (SC)Tillis, Thom (NC)

Toomey, Patrick J. (PA)Vitter, David (LA)

Voinovich, George Victor (OH)

Warner, John W. (VA)

Wicker, Roger F. (MS)Young, Todd (IN)

106 107 108 109 110 111 112 113 114 115

Congress Number

Sen

ator

s (R

epub

lican

)

Figure 6: Churning in Committee Membership: This figure depicts the frequency of com-mittee membership changes for each senator. Red (Blue) cell indicates that the senator moved toat least one (no) new committee in the congress that he/she did not serve in the previous congress.The white cell denotes the congress that the politician did not serve.

to belong to firm i’s connections: that is, she/he needs to represent the state where i has its

headquarter. Finally, the committee into which the politician enters, or from which it exits, must

have a non-zero weight for i. We illustrate this with Figure 7. It shows the returns of lobbying to

three firms when their own “connected” politicians change committee memberships in two periods.

Specifically, we suppose that P1 moves from Red (middle) to Blue (top) committee; P2 changes her

membership to Gray (bottom) committee; and P3 stays in Red committee. The color of committee

represents the most valuable committee for F1, F2, and F3 with the same boundary color. Firms

and politicians with the same shape (e.g., F1 and P1) are assumed to be politically connected. We

assume that the change of committee membership affects the value of lobbying. For example, F1’s

lobbying is expected to have higher returns than before when the politician that it has a closer

23

F1

F2

F3

Firm

Higher

Lower

Unchanged

Returns of Lobbying

P1

time t time t+ 1

Committee Membership

P1 P2 P3

P2

P3

w1

w2

w3

Figure 7: The Effects of Committee Membership Changes on Values of Lobbying: Thisfigure illustrates the identification strategy employed in the empirical analysis. It shows the returnsof lobbying when three politicians (P1, P2, and P3) who served in Red committee (middle) at timet change their committee memberships at t+ 1. Specifically, P1 moves from Red (middle) to Blue(top) committee; P2 changes her membership to Gray (bottom) committee; and P3 stays in Redcommittee. The color of committee represents the most valuable committee for F1, F2 (red), andF3 with the same boundary color. Firms and politicians with the same shape (e.g., F1 and P1) areassumed to be politically connected. We assume that the change of committee membership affectsthe value of lobbying. For example, F1’s lobbying is expected to have higher returns than beforewhen the politician that it has a closer tie to (i.e., P1) moves to the committee that it values. Incontrast, the value of lobbying would decrease for F2 when its connected politician leaves its mostvaluable red committee.

tie to (i.e., P1) moves to the committee that it values. In contrast, the value of lobbying would

decrease for F2 when its connected politician leaves its most valuable red committee.

Identification We discuss three potential challenges to the identification strategy, each one

related to each ingredient of the instrument. The first is related to the locations of firms and

politicians. If it were easy for any of the two to change locations over time, then this would threaten

the identification. For example, if firms can freely move to other states with representatives serving

in the committees that are highly relevant for them, then changes in committee membership

24

would directly influence firms’ location as well as their political connections, undermining the

identification. This is highly unlikely, because firms’ locations are often fixed before the changes

in committee membership that we exploit. Moreover, we do not see changes in firms’ headquarter

locations over time in our dataset. Similarly, the likelihood of a politician changing his/her district

is less than 1 percent. A second potential challenge to identification is that committee weights

could reflect anticipated changes in committee membership. In particular, if firms anticipate

changes in committee membership, then the timing of those changes will not be well identified.

We test this by evaluating the cross-section correlation between weights in t − k and changes in

committee membership in t. We find a correlation near zero.

The final issue is whether firms can directly influence the assignment of politicians into commit-

tees. We confirm that this does not hold because those decisions are determined by various factors

exogenous to firms such as electoral outcomes and inter-party negotiations, party’s independent

committee (e.g., Democrats’s Steering and Outreach Committee), and seniority20. To be sure,

firms may still indirectly influence the committee assignment. That is, committee membership

changes might be endogenous to firm characteristics and influence as politicians may select into cer-

tain committees in order to deliver targeted benefits to their politically connected firms. Although

it is certainly true that a politician’s “wish list”, reflecting the interests of their constituencies,

plays an important role in the committee assignment process, we emphasize that our identification

comes from the changes in the lobbying value over time. For example, Kamala Harris (D-CA) was

appointed to the Committee on Commerce, Science, and Transportation in the 115th Congress,

which might be endogenous to the importance of technology sector in California. However, we

focus on investigating how this new assignment increased the marginal value of lobbying especially

for the technology companies as no senators from the state served in the committee in the 114th

Congress. On the other hand, if senators from a certain state do not change their memberships

(e.g., Montana senators serving in the Agriculture, Nutrition, and Forestry Committee for their

beef sector), then such observations will not contribute to our estimation.21 In fact, politicians

often have to represent heterogeneous interests of their constituencies, and therefore the churning

of memberships that we observed in Figure 6 would be inconsistent with the presumption that

20See Schneider (2006) for further details about committee assignment process21Even in this case, we find that no senator from Montana served in the committee in the 114th Congress.

25

Log Sales Log Profits

(1) (2) (3) (4) (5) (6) (7) (8)

Log Lobby 0.0436∗∗∗ 0.0500∗∗∗ 0.319∗∗∗ 0.227∗∗∗ 0.0241∗∗ 0.0468∗∗∗ 0.322∗∗∗ 0.236∗

(0.00891) (0.0117) (0.0754) (0.0795) (0.00948) (0.0130) (0.114) (0.124)

Firm and Year FE X X X X X X X X

State-Year FE X X X X X X X X

Model OLS OLS IV IV OLS OLS IV IV

F-Stat 21.30 20.20 14.30 13.10

Sample All Post 2007 All Post 2007 All Post 2007 All Post 2007

Weight Lag t-1 t-1 t-1 t-1

N 15332 8796 15332 8796 10369 6064 10369 6064

Table 2: Firm Sales, Profits and Lobbying: This table presents the OLS and IV betweenlobbying expenditures and firms’ sales and profits. Profits are defined as sales minus wage bills,capital expenditures and intermediate input expenditures. All regressions have firm, year andstate-year fixed effects. The weights of the instrument are defined at t − 1. Standard errors aredouble clustered at firm and year level. *** p<0.01, ** p<0.05, * p<0.1

politicians select into certain committees and stay there always representing specific interests with

no variation over time.

Results Table 2 presents the main result of IV approach. Column 1 and 2 shows the simple

OLS between firms’ sales and lobbying expenditure, similar to the one presented in Figure 2, but

including a set of fixed effects. It shows that the correlation is significant and robust. Given the

endogeneity concerns, column 3 and 4 shows the IV’s second stage. It shows that the relationship

is significantly positive and bigger than the OLS. Taking column 4 as our preferred estimate, it

shows that an increase in 1 percent of lobbying expenditures translate into 0.2 percent increase

in revenues. The table shows that the F-stat of the first stage is sufficiently big. Finally, our

finding is robust to using firms profits as outcome, which takes into account factor expenditures

of firms such as labor, capital and intermediate inputs. More robustness to this specification can

be found in Appendix F. Given this strong causal relationship between lobbying expenditure and

sales, we proceed to the structural estimation to evaluate how important this relationship is for

the misallocation of resources and aggregate productivity.

26

4.2 Structural Estimation

In this section we describe the structural estimation of the model described in Section 3.1. The

estimation proceeds in three steps. First, a set of parameters are defined exogenously. A second set

of parameters are calibrated directly to analytical solutions of the model. Finally, the remaining

parameters are estimated via a simulated method of moments (SMM) procedure. We describe

each step in turn.

Exogenous Parameter Restrictions A set of parameters are set exogenously. First, given

that we do not have enough power to estimate heterogeneous δs’s, we set δs = δ for all s. Second,

given that we do not have sufficiently good data to estimate σCs , we set σCs = 4, a value in the

range of the values used in the literature (Hsieh and Klenow, 2009).22 Third, it is standard in this

literature that, given the free entry condition, the entry costs can be normalized to one. Fourth,

the death rate is taken from the literature and set to ηs = 0.025 (Bernard et al., 2007). Finally, we

assume a joint log-normal distribution for F , and assume that its mean is zero. This is without

loss of generality since it is straightforward to show that the model is invariant to these means. For

simplicity, we assume for now that this distribution is same across sectors. It is straightforward

to extend this to heterogeneous distributions across sectors.

Calibrated Parameters A set of parameters can be obtained directly from analytical solutions

of the model. First, θs are value added shares of sectors relative to total gross domestic product

(GDP). Second, αNs and αXs are labor and intermediate input costs relative to gross output.

Finally, given the assumption of CRS, we have that αKs = 1 − αLs − αXs . These moments can be

directly extracted from the data using information from the Bureau of Economic Analysis (BEA).

Both the results behind this calibration and the data used are standard in the literature (Hsieh

and Klenow, 2009; Caliendo et al., 2017). The moments for the value added shares of sectors, and

the Cobb-Douglas weights are shown in Figures 8, and 9 – 11.

Simulated Method of Moments Given the parameters set exogenously and calibrated from

analytical relationships in the model, the remaining parameters are estimated via a simulated

method of moments. This method is chosen given that the model does not have an analytical

solution of some parameters as a function of data. These parameters are the fixed costs fPs , fLs ,22We show how sensitive our results are to this assumption.

27

Val

ue A

dded

Sha

re0.

000.

020.

040.

060.

080.

100.

120.

14

Real e

stat

e an

d re

ntal

and

leas

ing

Man

ufac

turin

g

Profe

ssio

nal,

scie

ntific

& te

chni

cal s

ervic

es

Fina

nce

and

insu

ranc

e

Health

car

e an

d so

cial a

ssist

ance

Retai

l tra

de

Who

lesa

le tr

ade

Info

rmat

ion

Const

ruct

ion

Admin

and

was

te m

anag

emen

t ser

vices

Tran

spor

tatio

n an

d war

ehou

sing

Accom

mod

atio

n an

d fo

od s

ervic

es

Oth

er s

ervic

es, e

xcep

t gov

ernm

ent

Min

ing

Utilitie

s

Agricu

lture

, for

estry

, fish

ing

& hun

ting

Educa

tiona

l ser

vices

Arts, e

nter

tain

men

t & re

crea

tion

Figure 8: Value Added Share: This figure presents value added shares relative to total valueadded across sectors, for each sector and year, averaged across the period of 2000-2017. Owncalculations using data from the BEA. It corresponds to θsSs=1 in the model.

Labo

r S

hare

(re

lativ

e to

Gro

ss O

utpu

t)0.

00.

10.

20.

30.

40.

50.

6Edu

catio

nal s

ervic

es

Health

car

e an

d so

cial a

ssist

ance

Admin

and

was

te m

anag

emen

t ser

vices

Profe

ssio

nal,

scie

ntific

& te

chni

cal s

ervic

es

Oth

er s

ervic

es, e

xcep

t gov

ernm

ent

Retai

l tra

de

Accom

mod

atio

n an

d fo

od s

ervic

esW

hole

sale

trad

e

Arts, e

nter

tain

men

t & re

crea

tion

Const

ruct

ion

Tran

spor

tatio

n an

d war

ehou

sing

Fina

nce

and

insu

ranc

eIn

form

atio

nM

anuf

actu

ring

Utilitie

sM

inin

g

Agricu

lture

, for

estry

, fish

ing

& hun

ting

Real e

stat

e an

d re

ntal

and

leas

ing

Figure 9: Labor Share: Labor expenditures relative to gross output, for each sector and year, av-eraged across the period of 2000-2017. Own calculations using data from the BEA. It correspondsto αLs Ss=1 in the model.

the variances and covariances of the distribution F , and the returns to lobbying, δ. Thus, we

estimate the following vector of parameters:

Θ = fPs , fLs , v(φP ), v(φL), v(φτ ), cov(φP , φL), cov(φP , φτ ), cov(φL, φτ ), δ.

The algorithm proceeds in four steps. In the first, the model is simulated given a value for Θ.

Second, with the simulation of the model, a set of moments is produced and stacked into the

vector f(Θ). Third, the same set of moments is produced with data and stacked into the vector

f . Finally, an objective function is computed to evaluate the deviations of the simulated moments

28

Inte

rmed

iate

Inpu

t Sha

re0.

00.

10.

20.

30.

40.

50.

60.

7M

anuf

actu

ring

Agricu

lture

, for

estry

, fish

ing

& hun

ting

Tran

spor

tatio

n an

d war

ehou

sing

Const

ruct

ion

Fina

nce

and

insu

ranc

eIn

form

atio

n

Accom

mod

atio

n an

d fo

od s

ervic

esUtili

ties

Arts, e

nter

tain

men

t & re

crea

tion

Health

car

e an

d so

cial a

ssist

ance

Educa

tiona

l ser

vices

Oth

er s

ervic

es, e

xcep

t gov

ernm

ent

Profe

ssio

nal,

scie

ntific

& te

chni

cal s

ervic

es

Admin

and

was

te m

anag

emen

t ser

vices

Min

ing

Retai

l tra

de

Who

lesa

le tr

ade

Real e

stat

e an

d re

ntal

and

leas

ing

Figure 10: Intermediate Input Share: Intermediate input expenditures relative to gross output,for each sector and year, averaged across the period of 2000-2017. Own calculations using datafrom the BEA. It corresponds to αXs Ss=1 in the model.

Cap

ital S

hare

(re

lativ

e to

Gro

ss O

utpu

t)0.

00.

10.

20.

30.

40.

50.

60.

7

Real e

stat

e an

d re

ntal

and

leas

ing

Min

ing

Utilitie

s

Who

lesa

le tr

ade

Info

rmat

ion

Agricu

lture

, for

estry

, fish

ing

& hun

ting

Retai

l tra

de

Arts, e

nter

tain

men

t & re

crea

tion

Fina

nce

and

insu

ranc

e

Accom

mod

atio

n an

d fo

od s

ervic

esCon

stru

ctio

n

Tran

spor

tatio

n an

d war

ehou

sing

Oth

er s

ervic

es, e

xcep

t gov

ernm

ent

Profe

ssio

nal,

scie

ntific

& te

chni

cal s

ervic

esM

anuf

actu

ring

Admin

and

was

te m

anag

emen

t ser

vices

Health

car

e an

d so

cial a

ssist

ance

Educa

tiona

l ser

vices

Figure 11: Capital Share: Defined as the residual of labor and intermediate input shares, relativeto gross output, so that the condition of constant returns to scale hold: αKs = 1 − αLs − αXs .Computed for each sector and year, averaged across the period of 2000-2017. Own calculationsusing data from the BEA.

from the data moments, d(Θ) = f − f(Θ). If this difference is not minimized according to some

threshold, the algorithm is repeated for a different set of parameter values, until a minimum is

reached. The estimation procedure is based on the following moment condition:

E [d(Θ0)] = 0,

where Θ0 is the true value of Θ. Thus, the algorithm looks for Θ such that

Θ = argminΘd(Θ)′Wd(Θ),

where W is a weighting matrix which is the generalized inverse of the estimated variance-covariance

29

matrix of the moments calculated from the data.23

Moments Used and Related Parameters Three sets of moments are targeted in the data

to estimate the parameters of the model. Although the SMM procedure estimates all parameters

in Θ jointly, when presenting each set of moments we discuss intuitively how each moment used

is related to the parameters estimated. The first set involves the share of firms that lobby and

the distribution of the number of firms across sectors which are reported in Table 1. These

moments are related to the fixed cost of production and lobbying. The fixed cost are related to

the distribution of number of firms across sectors and the fixed lobbying cost to the share of firms

that select into lobbying activity. The second set involves moments of the joint distribution of firm

size, distortions, and lobbying expenditures. These moments are directly related to the matrix

of idiosyncratic primitives in the cross-section. More dispersion in the productivity primitive will

induce more dispersion in firm sales. Similarly for lobbying expenditures and firms’ output wedges.

The correlation between these variables is related in turn to the correlation between the primitives.

The third set of moments are the OLS and IV regressions of sales on lobbying expenditures

documented in Table 2. This set of moments relates to the returns to lobbying and the correlation

between lobbying productivity and lobbying expenditures. As is standard, the departures between

the true parameter in a regression and the biased one created by endogeneity issues is given by the

correlation between the endogenous variable and the residual in the regression. In our case, the

structural regression is given by Equation (6) in which the true parameter is (1− δ) and the bias

comes from the fact that lobbying expenditures are a function of the residual, which is lobbying

productivity φL. Thus, assuming that the identification assumption on the instrument holds, the

IV strategy correctly identifies (1− δ). Given that, the biased OLS estimate is the sum of the IV

estimate and the correlation between lobbying expenditures and lobbying productivity. This, in

turn is related to the parameters that govern the correlation between lobbying productivity and

other firms’ primitives that underly lobbying expenditures. Thus, targeting the estimate of the

second stage of both the OLS and IV model contributes to identifying both δ and the correlation

between firms’ primitives.

Estimation Result The results of the estimation are presented in Table 3. It shows that the

model is able to replicate the main features of the moments used from the data. In particular, in

23For now, we assume the identity matrix, which effectively weights all the moments equally.

30

Parameter Parameter Name Targeted Moment Data Model

fPs Production fixed cost Distribution of Number of Firms Table 1 Figure D.1

fLs Lobbying fixed cost Share of lobbying Firms Table 1 Figure D.2

v(φPs ) Variance of Production Productivity Firms’ Sales Dispersion 2.8 2.4

v(φLs ) Variance of Lobbying Productivity Firms’ Lobbying Expenditure Dispersion 2.1 1.8

v(φτs) Variance of Residual Distortions Firms’ Output Wedge Dispersion 1.1 1.6

corr(φPs , φLs ) Cov of Production and Lobbying Productivity Firms’ Correlation of Sales and Lobbying 0.5 0.3

corr(φPs , φτs) Cov of Production Productivity & Residual Distortions Firms’ Corr. of Sales and Output Wedges -0.5 -0.1

corr(φLs , φτs) Cov of Lobbying Productivity & Residual Distortions Firms’ Corr. of Lobbying and Output Wedges 0.2 0.4

βOLS Returns to Lobbying Biased OLS of Returns to Lobbying (1− δ) 0.05 0.045

βIV Returns to Lobbying IV Returns to Lobbying (1− δ) 0.23 0.19

Table 3: Parameter and Moments from the SMM: This table documents the results of theSMM procedure. It shows, for each parameter, the point estimate and the targeted moment.Note that column 4 and 5 of the last 2 rows show the reduced-form coefficient of the OLS and IVestimate, which is 1− δ.

order to generate a downward bias of the OLS estimate relative to the IV, it predicts a positive

correlation between firms’ production and lobbying productivity and a positive correlation between

firms’ lobbying productivity and residual distortion φτ . Similarly, the share of firms across sectors

and the share of firms within sectors that lobby are also well approximated by the estimated

model. With these results, we proceed to study different counterfactuals of how lobbying influences

aggregate productivity.

Counterfactual with no Lobbying We evaluate quantitatively how aggregate productivity

changes with lobbying activity. To understand the effect of lobbying activity, we consider a

counterfactual where δ = 0, i.e., firms choose endogenously not to lobby. In that case we find that

aggregate productivity would be 22 percent higher than the one where firms obtain the return to

lobbying that we estimate from the data.24 The left panel of Figure 12 shows sensitivity of how

those losses from vary with δ. The figure shows a slightly non-linear negative relationship between

productivity changes and the returns to lobbying. There are two main forces behind the losses of

lobbying activity. As Proposition 2 shows, one is because lobbying directly affects firms’ wedges,

which affect the dispersion of TFPR and thus how firms’ productivity is aggregated. This is the

traditional channel studied in Hsieh and Klenow (2009). The second channel is that by changing

24Note that, as is standard in this literature, we focus on aggregate productivity instead of aggregate output

since our theory does not include the accumulation of physical or human capital.

31

0.0 0.2 0.4 0.6 0.8 1.0

−0.

25−

0.20

−0.

15−

0.10

−0.

050.

00

β

Agg

rega

te P

rodu

ctiv

ity P

erce

ntag

e C

hang

es

2 3 4 5 6

−0.

35−

0.30

−0.

25−

0.20

−0.

15

σC

Agg

rega

te P

rodu

ctiv

ity P

erce

ntag

e C

hang

es

Figure 12: Simulation Results: This figure shows counterfactual analysis of the effect of lob-bying activity on aggregate productivity. The left panel shows percentage changes in aggregateproductivity for different values of δ, relative to the case with no lobbying, δ = 0. The rightpanel shows how changes in aggregate productivity when δIV = 0.81, vary with different σC . Thedashed red line corresponds to the values of the benchmark estimated model.

the allocation of resources, demand for labor can change which in turn changes factor prices and

thus entry of firms. Changes in entry also affect aggregate productivity since the household has a

utility that features love for variety. Of the aforementioned 22 percent losses, around 41 percent are

due to changes in entry. The remaining is given by changes in the allocation of resources among

existing firms. This highlights that the effect of lobbying on changes in entry is an important

margin to consider when evaluating its aggregate impact.

Given that we do not estimate the elasticity of substitution and that it is an important param-

eter as Proposition 2 shows, we evaluate what role the parameter plays in these results. The right

panel of Figure 12 shows how the changes in the 22 percent losses from lobbying activity change

with the value of σC , relative to the benchmark level of σC = 4. One can see that if σC declines to

2, the losses would still be of around 38 percent, whereas if it increases to 6 it would go down until

around 13 percent. The intuition is that as σC increases, there is less scope for the monopolistic

competition to extract its residual demand and the household can substitute away more easily

whatever effect lobbying activity has. On the other hand, if there is less substitution and products

are more differentiated, then lobbying has a bigger effect because the household cannot substitute

varieties when lobbying affects their prices. This shows that the elasticity of substitution in final

32

demand is important in governing the effects of lobbying on aggregate productivity.

5 Conclusions

This paper examines how firms’ lobbying activity in the U.S. affects aggregate productivity by

misallocating resources between firms. To address this, we went from micro causal estimates

of the effect of firms’ lobbying expenditures on firm size using a novel instrument, to aggregate

productivity by developing a heterogeneous firm model with endogenous lobbying. By estimating

the model with micro data, we show that firms’ lobbying activity decreases aggregate productivity

by 22 percent relative to an economy without lobbying activity. The main force behind this effect

is changes in the size distribution of firms because, through lobbying, some firms get bigger than

what they should. To the best of our knowledge, we are the first to evaluate quantitatively how

lobbying activity affects the aggregate misallocation of resources by distorting firms’ size. Thus,

by providing direct micro evidence and aggregate quantifications, we explored the understanding

of how firms’ political activities affects the allocation of resources in the economy.

Several open questions appear after these results and remain for future research. On the one

hand, what is dampening competition in these lobbying markets? Why are there not more firms

entering into lobbying activity? Why is there so much persistence in this lobbying? These are key

questions to improve the functioning of this market. On the other hand, how does the opposite

direction of the relationship between size and lobbying works? How does the business cycle of the

economy affect firms’ lobbying activity? This is also necessary if one is to fully understand how

the politics interact with the economics of firm behavior, and how this in turn impacts aggregate

productivity of an economy.

33

References

Adao, R., M. Kolesar, and E. Morales (2018). Shift-share designs: Theory and inference. Technical

report, National Bureau of Economic Research.

Ansolabehere, S., J. M. de Figueiredo, and J. M. Snyder (2003). Why is there so little money in

u.s. politics? The Journal of Economic Perspectives 17 (1), 105–130.

Ansolabehere, S., J. M. Snyder, and M. Tripathi (2002). Are PAC contributions and lobbying

linked? new evidence from the 1995 lobby disclosure act. Business and Politics 2, 131–155.

Arayavechkit, T., F. Saffie, and M. Shin (2018). Capital-based corporate tax benefits: Endogenous

misallocation through lobbying. Working Paper.

Baqaee, D. R. and E. Farhi (2017). Productivity and misallocation in general equilibrium. Tech-

nical report, National Bureau of Economic Research.

Bernard, A. B., S. J. Redding, and P. K. Schott (2007, 01). Comparative advantage and hetero-

geneous firms. The Review of Economic Studies 74 (1), 31–66.

Bertrand, M., M. Bombardini, R. Fisman, and F. Trebbi (2018). Tax-exempt lobbying: Corporate

philanthropy as a tool for political influence. Technical report, National Bureau of Economic Re-

search. available from https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3095686.

Bertrand, M., M. Bombardini, and F. Trebbi (2014). Is it whom you know or what you know?

an empirical assessment of the lobbying process. The American Economic Review 104 (12),

3885–3920.

Boehm, J. and E. Oberfield (2018). Misallocation in the market for inputs: Enforcement and the

organization of production. Technical report, National Bureau of Economic Research.

Bombardini, M. (2008). Firm heterogeneity and lobby participation. Journal of International

Economics 75 (2), 329 – 348.

Bombardini, M. and F. Trebbi (2012). Competition and political organization: Together or alone

in lobbying for trade policy? Journal of International Economics 87 (1), 18–26.

34

Caliendo, L., F. Parro, E. Rossi-Hansberg, and P.-D. Sarte (2017). The impact of regional and

sectoral productivity changes on the us economy. The Review of economic studies 85 (4), 2042–

2096.

De Figueiredo, J. M. and B. K. Richter (2014). Advancing the empirical research on lobbying.

Annual Review of Political Science 17, 163–185.

Fouirnaies, A. and A. B. Hall (2017, 2018/10/28). How do interest groups seek access to commit-

tees? American Journal of Political Science 62 (1), 132–147.

Gawande, K. and U. Bandyopadhyay (2000). Is protection for sale? evidence on the grossman-

helpman theory of endogenous protection. The Review of Economics and Statistics 82 (1), pp.

139–152.

Goldberg, P. K. and G. Maggi (1999). Protection for sale: An empirical investigation. The

American Economic Review 89 (5), pp. 1135–1155.

Goldsmith-Pinkham, P., I. Sorkin, and H. Swift (2018). Bartik instruments: What, when, why,

and how. National Bureau of Economic Research Working Paper Series.

Grossman, G. M. and E. Helpman (1994, 09). Protection for sale. The American Economic

Review 84 (4), 833–850.

Hall, R. L. and A. V. Deardorff (2006). Lobbying as legislative subsidy. American Political Science

Review 100 (1), 69–84.

Hall, R. L. and F. W. Wayman (1990). Buying time: Moneyed interests and the mobilization of

bias in congressional committees. American political science review 84 (3), 797–820.

Haltiwanger, J., R. Kulick, and C. Syverson (2018). Misallocation measures: The distortion that

ate the residual. Working Paper.

Hansen, W. L. and N. J. Mitchell (2000). Disaggregating and explaining corporate political

activity: Domestic and foreign corporations in national politics. American Political Science

Review 94 (4), 891–903.

35

Hsieh, C.-T. and P. J. Klenow (2009, 11). Misallocation and manufacturing tfp in china and india.

The Quarterly Journal of Economics 124 (4), 1403–1448.

Kang, K. (2015). Policy influence and private returns from lobbying in the energy sector. The

Review of Economic Studies 83 (1), 269–305.

Khwaja, A. I. and A. Mian (2005, 11). Do lenders favor politically connected firms? rent provision

in an emerging financial market. The Quarterly Journal of Economics 120 (4), 1371–1411.

Kim, I. S. (2017). Political cleavages within industry: Firm-level lobbying for trade liberalization.

American Political Science Review 111 (1), 1–20.

Kim, I. S. (2018). Lobbyview: Firm-level lobbying & congressional bills database. Working Paper

available from http://web.mit.edu/insong/www/pdf/lobbyview.pdf.

Kim, I. S. and I. Osgood (2018). Firms in trade and trade politics. Annual Review of Political

Science, Forthcoming.

Melitz, M. J. (2003, 11). The impact of trade on intra-industry reallocations and aggregate

industry productivity. Econometrica 71 (6), 1695–1725.

Midrigan, V. and D. Y. Xu (2014, 02). Finance and misallocation: Evidence from plant-level data.

The American Economic Review 104 (2), 422–458.

Moll, B. (2014). Productivity losses from financial frictions: Can self-financing undo capital

misallocation? American Economic Review 104 (10), 3186–3221.

Powell, E. N. and J. Grimmer (2016, 2018/10/28). Money in exile: Campaign contributions and

committee access. The Journal of Politics 78 (4), 974–988.

Restuccia, D. and R. Rogerson (2008). Policy distortions and aggregate productivity with hetero-

geneous establishments. Review of Economic Dynamics 11 (4), 707–720.

Richter, B. K., K. Samphantharak, and J. F. Timmons (2009). Lobbying and taxes. American

Journal of Political Science 53 (4), 893–909.

36

Schneider, J. (2006). Committee assignment process in the us senate: Democratic and republican

party procedures. Congressional Research Service Report for Congress.

Stewart, C. and J. Woon (2011). Congressional committee assignments, 103rd to 112th congresses,

1993–2011. available from http://web.mit.edu/~17.251/www/data_page.html.

Tullock, G. (1972). The purchase of politicians. Western Economic Journal 10 (3), 354–55.

37

Appendix A Construction of Lobbying Dataset

Firm’s lobbying activity is built from public reports from the SOPR. These reports are required

to be filled by any lobbyist in the US due to the Lobbying Disclosure Act of 1995. Lobbyists must

file 3 types of reports depending on their activity, i.e., LD-1, LD-2 and LD-203.25 The LD-1 form

contains information about registrants, i.e., lobbyists, and clients such as their name, address, and

principal place of business. The LD-203 form presents the disclosure of all political committees

established or controlled by a lobbyist and all federal campaign contributions of $200 or more.

Finally, the LD-2 form is the reporting form where registrants disclose their lobbying activities

and related expenses. Dollar amounts of lobbying reported in section 12 and 13 are estimates of

income (lobbyists) or expenses (in-house lobbying) spent in the reporting period rounded to the

nearest $5,000. When total amount is less than $5,000, registrants should still file a report and

include a statement indicating the fact.26 In addition to the general issues categories, it is legally

required that registrants report any congressional bills numbers they have lobbied as well as the

description of their activities in section 16. An example LD-2 report can be found in Appendix B.

We use lobbying information available from all LD-1 and LD-2 reports filed between 1999 and

2015 under the Lobbying Disclosure Act of 1995 (amended by the Honest Leadership and Open

Government Act of 2007).

Since the reports are in documents that are not directly manageable to use for empirical

research, there are several steps necessary to be able to use the information in them. We first

directly parse the reports to build a report-level dataset. In doing so, each report is carefully

examined whether there exists any amendments, and if so only the latest report is kept based

on the date and time of filing. This is an important step because researchers will erroneously

overweight firm’s lobbying activity by duplicating multiple reports with essentially similar contents

and lobbying expenses.27

25All filings are updated quarterly in a digitized compressed XML format. As of Septem-ber 2015, there are more than 1 million LD-1 and LD-2 reports publicly available fromhttp://www.senate.gov/legislative/Public Disclosure/database download.htm.

26 We note that registrants were required to file reports biannually (instead of quarterly) prior to the HonestLeadership and Open Government Act of 2007 amendment. Before 2008, estimates of amounts in excess of $10,000was rounded to the nearest $20,000. We address this difference by considering firm-year as the unit of analysisafter aggregating quarterly or biannual reports for a given year.

27We note that no empirical study, using the lobbying reports either from SOPR or fromhttp://www.opensecrets.org/, has discussed this problem to the best of our knowledge. Thus, we suspect that

1

We then create a mapping from clients to their unique identifiers in databases such as Com-

pustat and Orbis (Bureau van Dijk) allowing us to link firm’s economic characteristics to their

political behavior. Finding a unique firm identifier is challenging because the matching can be

done only through client names (i.e., character strings) which tend to exist in many different for-

mats even for the same firm. For example, Apple Inc. appears in 15 different client names: APPLE

INC, Apple, Inc., Apple, Apple Inc., Apple Inc, APPLE COMPUTER, INC., APPLE, Apple, Inc,

APPLE COMPUTERS, APPLE COMPUTER, APPLE COMPUTERS, INC, APPLE COMPUTER INC, APPLE INC.,

APPLE COMPUTERS INC, APPLE COMPUTER, INC. Although some of these can be easily addressed

by removing dot and suffix, in many cases it is not straightforward to distinguish misspelled client

names and abbreviations from their legal firm names. To address this problem, we employ four

strategies. First, we use FuzzyWuzzy string matching algorithm comparing the full list of public

firm names from Compustat against 61,478 unique client names.28 Second, we use Bureau van

Dijk server’s Batch Search functionality to find each firm’s ISIN and ticker symbol, which will

then be used to find Compustat identifier code of clients.29 Third, we use Center for Respon-

sive Politics lobbying data to check whether any additional matching can be achieved by using

their Standardized client variable. Finally, we randomly sample 5% of client names to verify

whether any publicly trading firms were missed so that we can improve the matching algorithm

from the first step. We update our matching algorithm quarterly each time a new set of reports

become available. This process ends up with a database at the report level that has 972,005 ob-

servations. Each observation contains a report id, the id of the lobbyist, the total amount lobbied,

whether lobbying activity was outsourced or not, all the issues lobbied, and the bill number if

the information is available. For reports that are filled by Compustat firms, we have the unique

identifier of Compustat firms and all the information given by Compustat.

most of the existing studies might contain numerous duplicates by including both original filings and their amend-ments.

28We use the following natural language processing module from Python programming language:https://pypi.python.org/pypi/fuzzywuzzy

29 Unfortunately, the batch search can be conducted only on 1,000 firm names each time. Thus, we repeated thequeries more than 60 times to get the full search results.

2

Appendix B LD-2 Report Example

Clerk of the House of Representatives Legislative Resource Center B-106 Cannon Building Washington, DC 20515 http://lobbyingdisclosure.house.gov

Secretary of the Senate Office of Public Records 232 Hart Building Washington, DC 20510 http://www.senate.gov/lobby LOBBYING REPORT

Lobbying Disclosure Act of 1995 (Section 5) - All Filers Are Required to Complete This Page

1. Registrant Name Organization/Lobbying Firm Self Employed IndividualCapitol Tax Partners, LLP

2. AddressAddress1 101 Constitution Avenue, NW Suite 675 East Address2

City Washington State DC Zip Code 20001 Country USA

3. Principal place of business (if different than line 2)City State Zip Code Country

4a. Contact Name b. Telephone Number c. E-mail Mr. Christopher Javens 2022898700 faddoul@capitoltax.com

5. Senate ID# 65976-12

7. Client Name Self Check if client is a state or local government or instrumentality

Apple6. House ID# 356170002

TYPE OF REPORT 8. Year 2018 Q1 (1/1 - 3/31) Q2 (4/1 - 6/30) Q3 (7/1 - 9/30) Q4 (10/1 - 12/31)

9. Check if this filing amends a previously filed version of this report 10. Check if this is a Termination Report Termination Date 11. No Lobbying Issue Activity

INCOME OR EXPENSES - YOU MUST complete either Line 12 or Line 1312. Lobbying 13. Organizations

INCOME relating to lobbying activities for this reporting period was: EXPENSE relating to lobbying activities for this reporting period were:

Less than $5,000 Less than $5,000

$5,000 or more $ 90,000.00 $5,000 or more $

Provide a good faith estimate, rounded to the nearest $10,000, of all lobbying related income from theclient (including all payments to the registrant by any other entity for lobbying activities on behalf of theclient).

14. REPORTING Check box to indicate expense accounting method. See instructions for description ofoptions.

Method A. Reporting amounts using LDA definitions only

Method B. Reporting amounts under section 6033(b)(8) of the Internal Revenue Code

Method C. Reporting amounts under section 162(e) of the Internal Revenue Code

Signature Digitally Signed By: Christopher Javens Date 4/19/2018 4:13:31PM

LOBBYING ACTIVITY. Select as many codes as necessary to reflect the general issue areas in which the registrant engaged in lobbying on behalf of the client during the reporting period. Using a separate page foreach code, provide information as requested. Add additional page(s) as needed.

15. General issue area code TAX

16. Specific lobbying issues

Matters dealing with International Taxation, and H.R. 1, Bill to provide for reconciliation pursuant to titles II and V of the concurrent resolution on the budget for fiscal year 2018 (bill formerly known as the Tax Cutsand Jobs Act).

17. House(s) of Congress and Federal agencies Check if None

U.S. SENATE, U.S. HOUSE OF REPRESENTATIVES, Treasury - Dept of, Executive Office of the President (EOP)

18. Name of each individual who acted as a lobbyist in this issue area

First Name Last Name Suffix Covered Official Position (if applicable) New

Jonathan Talisman

Christopher Javens

19. Interest of each foreign entity in the specific issues listed on line 16 above Check if None

Information Update Page - Complete ONLY where registration information has changed.

20. Client new address

Address City State Zip Code Country

21. Client new principal place of business (if different than line 20)

City State Zip Code Country

22. New General description of client’s business or activities

LOBBYIST UPDATE

23. Name of each previously reported individual who is no longer expected to act as a lobbyist for the client

First Name Last Name Suffix First Name Last Name Suffix

1 3

2 4

ISSUE UPDATE

24. General lobbying issue that no longer pertains

AFFILIATED ORGANIZATIONS

25. Add the following affiliated organization(s)

Internet Address:

Name

AddressPrincipal Place of Business (city and state or country)

Street AddressCity State/Province Zip Country

CityState Country

26. Name of each previously reported organization that is no longer affiliated with the registrant or client

1 2 3

FOREIGN ENTITIES

27. Add the following foreign entities:

Name

AddressPrincipal place of business (city and state or country)

Amount of contribution forlobbying activities

Ownershippercentage in client

Street AddressCity State/Province Country

CityState Country

%

28. Name of each previously reported foreign entity that no longer owns, or controls, or is affiliated with the registrant, client or affiliated organization

1 3 5

Figure B.1: Report by Apple Inc., first quarter in 2018: A report filed by Apple Inc. showsthat Capital Tax Partners, LLP lobbied on behalf of Apple Inc. to lobby on Taxation issue (Section15). In particular, it lobbied on the House bill H.R.1 titled “An Act to provide for reconciliationpursuant to titles II and V of the concurrent resolution on the budget for fiscal year 2018” (Section16).

3

Appendix C Supporting Facts

In this appendix we document a set of facts that support the analysis in the main text.

C.1 Distribution of the Number of Lobbying Clients

Issues Mean Median Minimum Maximum Total Number of Bills

Accounting 1.4 1 1 12 1,043Advertising 5.2 4.5 1 27 82Aerospace 2.6 2 1 15 76Agriculture 2.6 2 1 195 1,082

Alcohol.DrugAbuse 2.1 1 1 10 200Animals 2.1 2 1 11 412

ApparelInd 1.5 1 1 7 140Arts.Entertainment 1.9 1 1 7 42

Automotive.Ind 3.8 2 1 37 319Aviation 4.1 2 1 93 836

Bakruptcy 4.4 2 1 29 78Banking 3.0 2 1 45 1,646

BeverageInd 3.3 3 1 15 27Budget.Approp 8.2 2 1 421 2,577

Chemicals 4.8 2 1 83 124CivilRights 1.9 2 1 40 1,263

CleanAir.Water 5.7 2 1 104 1,289Commodities 8.1 3 1 35 35

Communications 3.0 2 1 52 757ComputerInd 4.1 2 1 24 255Constitution 1.6 1 1 9 141

ConsumerIssues 5.5 2 1 73 825Copyright 8.4 3 1 151 577Defense 6.3 2 1 149 985

DisasterPlanning 1.9 1 1 9 261DistrictofColumbia 1.9 2 1 9 29

Economics 2.1 1 1 18 191Education 1.9 1 1 32 2,825

Energy.Nuclear 6.0 3 1 328 2,780Environment 3.0 2 1 190 1,117FamilyIssues 1.6 2 1 15 726FinancialInst 4.2 2 1 338 1,404

Firearms 1.6 1 1 12 644FoodInd 3.1 2 1 72 560

ForeignRelations 1.6 1 1 19 1,322FuelGasOil 3.1 2 1 20 264Gambling 3.4 2 1 16 99

GovernmentIssues 2.0 1 1 139 2,399HealthIssues 3.1 2 1 319 6,797

HomelandSecurity 8.2 2 1 158 816Housing 2.2 2 1 35 722

Immigration 2.8 2 1 133 1,249IndianAffairs 1.6 1 1 11 495

Insurance 4.0 2 1 82 931Intelligence 4.0 3 1 17 58

LawEnforcement 1.8 1 1 30 1,172Manufacturing 2.4 1.5 1 25 90Marine.Boating 2.2 2 1 27 561

Media 3.1 2 1 17 53Medical 1.7 1 1 7 209Medicare 3.1 2 1 118 2,726

Minting.Money 1.6 1 1 6 55NaturalResources 2.1 2 1 41 1,661

Pharmacy 3.9 2 1 27 269Postal 3.2 1 1 29 211

Railroads 6.2 3 1 56 307RealEstate 1.6 1 1 5 502

Religion 1.3 1 1 3 93Retirement 3.3 2 1 134 1,062

Roads.Highway 4 2 1 49 71Science.Technology 2.7 2 1 30 400

SmallBusiness 1.8 1 1 10 483Sports.Athletics 1.7 1 1 6 59

TariffsMisc 1.7 1 1 30 1,655Taxation 4.4 2 1 491 5,940

Telecommunications 4.8 3 1 77 1,219Tobacco 3.6 3 1 24 222

Torts 5 2 1 64 237Trade 3.2 2 1 135 1,862

Travel.Tourism 3.2 2 1 17 95Trucking.Shipping 3.8 2 1 33 128

Unemployment 2.0 2 1 6 48UrbanDvlpmnt 1.8 1 1 9 152

utilities 3.8 2 1 29 185Veterans 1.5 1 1 16 2,664Waste 2.0 1 1 8 59

Welfare 1.5 1 1 7 97Total 3.5 2 1 491 65,047

Table C.1: This table shows that the skewed distribution that we observed in Figure 5 holds truefor various other issues. We categorize each bill based on the frequency of the bill’s appearanceunder particular issue codes across reports. Most bills are lobbied by one or two interest groups.

4

C.2 Changes in Committee Membership

This subsection highlights in more detail how committee membership changes over time for politi-

cians. Figure C.1 shows the likelihood of switching committees, for each politician and each

congress. Blue squares indicate that a politician did not change any committee membership be-

tween two congresses. As one can see, there are few politicians that never change their committee

membership, i.e., politicians that have only blue squares in their corresponding row. To understand

the quantitative meaning of this, Figure C.2 shows the likelihood of changing a committee over

time. It shows that this likelihood is around 30 percent on average across Congress. Furthermore,

it highlights that this number has been fairly constant over time.

Pro

port

ion

of P

oliti

cian

s w

ithC

omm

ittee

Mem

bers

hip

Cha

nges

0.0

0.2

0.4

0.6

0.8

1.0

103 104 105 106 107 108 109 110 111 112 113 114

Congress

Figure C.2: Changes in Committee Membership: This figure shows the proportion of politi-cians who changed his/her committee membership in the standing committees across congressionalsessions. It shows that about 29% of politicians change their memberships.

5

Akaka, Daniel K. (HI)Baldwin, Tammy (WI)

Baucus, Max (MT)Bayh, Evan (IN)

Begich, Mark (AK)Bennet, Michael (CO)

Biden, Joseph R., Jr. (DE)Bingaman, Jeff (NM)

Blumenthal, Richard (CT)Booker, Cory (NJ)

Boxer, Barbara (CA)Breaux, John B. (LA)Brown, Sherrod (OH)

Bryan, Richard H. (NV)Burris, Roland (IL)

Byrd, Robert C. (WV)Cantwell, Maria (WA)

Cardin, Benjamin L. (MD)Carnahan, Jean (MO)

Carper, Thomas Richard (DE)

Casey, Robert P., Jr. (PA)Cleland, Max (GA)

Clinton, Hillary Rodham (NY)Conrad, Kent (ND)

Coons, Christopher (DE)

Cortez Masto, Catherine (NV)

Corzine, Jon Stevens (NJ)

Cowan, William (Mo) (MA)

Daschle, Thomas A. (SD)Dayton, Mark (MN)

Dodd, Christopher, J (CT)Donnelly, Joe (IN)

Dorgan, Byron (ND)

Duckworth, Tammy (IL)Durbin, Richard J. (IL)

Edwards, John (NC)

Feingold, Russel D. (WI)Feinstein, Dianne (CA)

Franken, Al (MN)

Gillibrand, Kirsten E. (NY)Graham, Bob (FL)Hagan, Kay (NC)Harkin, Tom (IA)

Harris, Kamala (CA)Hassan, Maggie (NH)Heinrich, Martin (NM)Heitkamp, Heidi (ND)

Hirono, Mazie (HI)

Hollings, Ernest F. (SC)Inouye, Daniel K. (HI)

Jeffords, James M. (VT)Johnson, Tim (SD)

Kaine, Timothy (VA)Kaufman, Ted (DE)

Kennedy, Edward M. (MA)Kerrey, J. Robert (NE)

Kerry, John F. (MA)Kirk, Paul (MA)

Klobuchar, Amy (MN)Kohl, Herbert H. (WI)

Landrieu, Mary L. (LA)

Lautenberg, Frank R. (NJ)Leahy, Patrick J. (VT)

Levin, Carl (MI)

Lieberman, Joseph I. (CT)

Lincoln, Blanche Lambert (AR)Manchin, Joe (WV)

Markey, Edward J. (Ed) (MA)McCaskill, Claire (MO)

Menendez, Robert (NJ)Merkley, Jeff (OR)

Mikulski, Barbara A. (MD)Miller, Zell Bryan (GA)

Moynihan, Daniel P. (NY)

Murphy, Christopher (CT)Murray, Patty (WA)

Nelson, Clarence William (Bill) (FL)

Nelson, Earl Benjamin (Ben) (NE)Obama, Barack (IL)

Peters, Gary (MI)Pryor, Mark (AR)Reed, Jack (RI)

Reid, Harry (NV)Robb, Charles S. (VA)

Rockefeller, John D., IV (WV)

Salazar, Kenneth Lee (CO)

Sarbanes, Paul S. (MD)Schatz, Brian E. (HI)

Schumer, Charles Ellis (Chuck) (NY)

Shaheen, Jeanne (NH)Specter, Arlen (PA)

Stabenow, Deborah Ann (MI)Tester, Jon (MT)

Torricelli, Robert G. (NJ)Udall, Mark (CO)Udall, Tom (NM)

Van Hollen, Chris (MD)Walsh, John E. (MT)

Warner, Mark (VA)

Warren, Elizabeth (MA)Webb, Jim (VA)

Wellstone, Paul David (MN)

Whitehouse, Sheldon (RI)Wyden, Ron (OR)

106 107 108 109 110 111 112 113 114 115Congress Number (106th − 115th)

Sena

tors

(Dem

ocra

ts)

No Committee Switch Complete Switch

Abraham, Spencer (MI)

Alexander, Lamar (TN)Allard, Wayne (CO)Allen, George (VA)

Ashcroft, John (MO)Ayotte, Kelly (NH)

Barrasso, John A. (WY)

Bennet, Robert F. (UT)Blunt, Roy (MO)

Bond, Christopher S. (MO)

Boozman, John (AR)Brown, Scott (M

A)

Brownback, Samuel Dale (KS)

Bunning, James Paul David (KY)Burns, Conrad (MT)

Burr, Richard (NC)

Campbell, Ben Nighthorse (CO)

Capito, Shelley M. (WV)Cassidy, Bill (LA)

Chafee, John H. (RI)

Chafee, Lincoln Davenport (RI)

Chambliss, Saxby (GA)Chiesa, Jeffrey (NJ)

Coats, Dan (IN)

Coburn, Thomas Allen (OK)

Cochran, Thad (MS)

Coleman, Norm (MN)Collins, Susan (ME)

Corker, Bob (TN)Cornyn, John (TX)Cotton, Tom (AR)

Coverdell, Paul (GA)Craig, Larry E. (ID)

Crapo, Michael Dean (ID)Cruz, Ted (TX)

Daines, Steve (MT)

DeMint, James W. (SC)

DeWine, Michael (OH)

Dole, Elizabeth Hanford (NC)

Domenici, Pete V. (NM)

Ensign, John Eric (NV)

Enzi, Michael B. (WY)Ernst, Joni (IA)

Fischer, Deb (NE)

Fitzgerald, Peter G (IL)Flake, Jeff (AZ)

Frist, William H. (TN)Gardner, Cory (CO)Gorton, Slade (WA)

Graham, Lindsey O. (SC)Gramm, Phil (TX)Grams, Rod (MN)

Grassley, Charles E. (IA)Gregg, Judd (NH)

Hagel, Chuck (NE)

Hatch, Orrin G. (UT)Heller, Dean (NV)

Helms, Jesse (NC)Hoeven, John (ND)

Hutchinson, Tim (AR)

Hutchison, Kay Bailey (TX)

Inhofe, James M. (OK)

Isakson, Johnny (GA)

Jeffords, James M. (VT)Johanns, Mike (NE)Johnson, Ron (WI)

Kennedy, John N. (LA)Kirk, Mark (IL)Kyl, Jon (AZ)

Lankford, James (OK)Lee, Mike (UT)

LeMieux, George (FL)Lott, Trent (MS)

Lugar, Richard G. (IN)Mack, Connie (FL)

Martinez, Melquiades R. (Mel) (FL)McCain, John (AZ)

McConnell, Mitch (KY)Moran, Jerry (KS)

Murkowski, Frank H. (AK)

Murkowski, Lisa (AK)Nickles, Don (OK)

Paul, Rand (KY)Perdue, David (GA)Portman, Rob (OH)

Risch, James (ID)Roberts, Pat (KS)

Roth, William V., Jr. (DE)Rounds, Mike (SD)Rubio, Marco (FL)

Santorum, Rick (PA)Sasse, Ben (NE)

Scott, Tim (SC)Sessions, Jeff (AL)

Shelby, Richard C. (AL)

Smith, Gordon H. (OR)

Smith, Robert C. (NH)

Snowe, Olympia J. (ME)Specter, Arlen (PA)Stevens, Ted (AK)

Strange, Luther (AL)

Sullivan, Daniel (AK)

Sununu, John E. (NH)

Talent, James Matthes (MO)

Thomas, Craig (WY)

Thompson, Fred (TN)Thune, John (SD)

Thurmond, Strom (SC)Tillis, Thom (NC)

Toomey, Patrick J. (PA)Vitter, David (LA)

Voinovich, George Victor (OH)

Warner, John W. (VA)

Wicker, Roger F. (MS)Young, Todd (IN)

106 107 108 109 110 111 112 113 114 115

Congress Number (106th − 115th)

Sen

ator

s (R

epub

lican

s)

Figure C.1: Changes in Committee Membership: This figure distinguishes the degrees ofcommittee membership changes for democrats (left) and republicans (right), providing furtherdetails to Figure 6.

6

C.3 List of Standing Committees

Senate House

Agriculture, Nutrition, and Forestry Agriculture

Appropriations Appropriations

Armed Services Armed Services

Banking, Housing, and Urban Affairs Budget

Budget Education and the Workforce

Commerce, Science, and Transportation Energy and Commerce

Energy and Natural Resources Ethics

Environment and Public Works Financial Services

Finance Foreign Affairs

Foreign Relations Homeland Security

Health, Education, Labor, and Pensions House Administration

Homeland Security and Governmental Affairs Judiciary

Judiciary Natural Resources

Rules and Administration Oversight and Government Reform

Small Business and Entrepreneurship Rules

Veterans’ Affairs Science, Space, and Technology

Small Business

Transportation and Infrastructure

Veterans’ Affairs

Ways and Means

Table C.2: This table presents the list of standing committees in the Senate and the House thatwe consider in the analysis.

7

Appendix D Structural Estimation

This appendix shows the fit of the estimated version of the model to two sets of moments from

the data: the percentage of firms that lobby in each sector, and the distribution of the number of

firms across sectors. In both cases, the figures demonstrate a relatively good fit of the model.P

erce

ntag

e of

Fir

ms

05

1015

2025

30Con

stru

ctio

nM

anuf

actu

ring

Info

rmat

ion

Admin

and

was

te m

anag

emen

t ser

vices

Min

ing

Oth

er s

ervic

es, e

xcep

t gov

ernm

ent

Agricu

lture

, for

estry

, fish

ing

& hun

ting

Fina

nce

and

insu

ranc

e

Health

car

e an

d so

cial a

ssist

ance

Educa

tiona

l ser

vices

Profe

ssio

nal,

scie

ntific

& te

chni

cal s

ervic

esUtili

ties

Retai

l tra

de

Arts, e

nter

tain

men

t & re

crea

tion

Tran

spor

tatio

n an

d war

ehou

sing

Real e

stat

e an

d re

ntal

and

leas

ing

Accom

mod

atio

n an

d fo

od s

ervic

esW

hole

sale

trad

e

DataEstimated Moments

Figure D.1: Number of Firms Share Fit: This figure shows the distribution of the numberof firms across sectors, both in the data and the one simulated from the estimated version of themodel.

Per

cent

age

of L

obby

ing

Fir

ms

05

1015

2025

Real e

stat

e an

d re

ntal

and

leas

ing

Agricu

lture

, for

estry

, fish

ing

& hun

ting

Retai

l tra

de

Accom

mod

atio

n an

d fo

od s

ervic

es

Health

car

e an

d so

cial a

ssist

ance

Profe

ssio

nal,

scie

ntific

& te

chni

cal s

ervic

esCon

stru

ctio

n

Tran

spor

tatio

n an

d war

ehou

sing

Utilitie

s

Oth

er s

ervic

es, e

xcep

t gov

ernm

ent

Info

rmat

ion

Fina

nce

and

insu

ranc

e

Arts, e

nter

tain

men

t & re

crea

tion

Admin

and

was

te m

anag

emen

t ser

vices

Educa

tiona

l ser

vices

Who

lesa

le tr

ade

Min

ing

Man

ufac

turin

g

DataEstimated Moments

Figure D.2: Lobbying Share Fit: This figure shows the percentage of firms in each sector thatlobby, both in the data and the one simulated from the estimated version of the model.

8

Appendix E Model

This appendix presents more details about the model described in the paper.

E.1 Model Derivations

The solution of the household problem is

cs(φ) = pys(φ)−σCs EsP

σCs −1s , (15)

where Ps =

[∫pys(φ)1−σCs MsdFs(φ)

] 1

1−σCsand Es = θsE is the price index and total expenditure

in sector s, respectively. Also, from sector-level optimization,

P = ΠSs=1

(Psθs

)θs(16)

and E is the aggregate price index and aggregate expenditure, respectively. Given the solution in

(15) and the fact that aggregate expenditures have to be equal to aggregate income, E = I , we

get that sectoral welfare is

Ys =EsPs

(17)

where Es = θs(wN + rK − T ).

The solution to firms’ optimization problem implies

pys(φ) =

µCsφP

(wαNs

)αNs ( rαKs

)αKs ( pXsαXs

)1−αNs −αKs, if φ∗s ≤ φ < φ∗∗s

11+τs(φ)

µCsφP

(wαNs

)αNs ( rαKs

)αKs ( pXsαXs

)1−αNs −αKs, if φ ≥ φ∗∗s

where µCs = σCsσCs −1

. Thus, The revenue function is rs(φ) = pys(φ)1−σCs EsPσCs −1s and thus,

rs(φ) ∝

(φPs )σCs −1, if φ∗s ≤ φ < φ∗∗s

(φPs (1 + τs(φ)))σCs −1, if φ ≥ φ∗∗s

9

Thus, the profit function is

πs(φ) =

rs(φ)σCs− wfPs , if φ∗s ≤ φ < φ∗∗s

(1 + τs(φ))rs(φ)

[1σCs− δs

(φLs l(φ))δs

1+τs(φ)

]− w(fPs + fLs ), if φ ≥ φ∗∗s .

Thus, when the firm is evaluating whether to lobby, it needs to compare the benefit given by

the policy with both the variable and the fixed cost.

Firms’ zero profit condition in (4) and (5) imply that

rs(φ∗s) = σCs wf

Ps (18)

(1 + τs(φ∗∗s ))rs(φ

∗∗s )

[1− σCs δs

(φL∗∗s l(φ∗∗s )

)δs1 + τs(φ∗∗s )

]− rs(φ∗∗s ) = σCs wf

Ls

Thus, we can write the revenue function as follows:

rs(φ) =

(φPsφP∗s

)σCs −1

σCs wfPs , if φ

∗s ≤ φ < φ∗∗s ,(

φPs (1+τs(φ))φP∗s

)σCs −1

σCs wfPs , if φ ≥ φ∗∗s ,

and the profit function as follows:

πs(φ) =

[(

φPsφP∗s

)σCs −1

− 1

]wfPs , if φ

∗s ≤ φ < φ∗∗s ,[

κs(φ)(φPsφP∗s

)σCs −1

− 1

]wfPs − wfLs , if φ ≥ φ∗∗s ,

where κs(φ) = (1+τ(φ))σCs

(1− σCs δs

(φLs l(φ))δs

1+τs(φ)

)is the factor that scales up profits relative to non-

lobbying profits, gross of the fixed cost of lobbying. Note that, in order to evaluate the benefits

of lobbying, the firm needs to compare the effect of κs(φ) against the fixed cost. This is the only

difference relative to the results in the closed economy version of the model in Melitz (2003).

The zero-profit condition of lobbying can be written as follows:

(κs(φ∗∗s )− 1)

(φP∗∗s

φP∗s

)σCs −1

=fLsfPs

(19)

10

Given selection into lobbying, average profits, conditional on successful entry can be expressed

as:

πs =(1− ξLs

)πNLs + ξLs π

Ls , (20)

where

πNLs =

∫ φ∗∗s

φ∗s

πs(φ)dF (φ)

F (φ∗∗s )− F (φ∗s),

πLs =

∫ ∞φ∗∗s

πs(φ)dF (φ)

1− F (φ∗∗s ),

ξLs =1− F (φ∗∗s )

1− F (φ∗s)(21)

where ξLs is the probability of lobbying in sector s. Average revenue, rs can be defined similarly.

Given selection into entry, the ex-post primitives distribution conditional on successful entry

is:

Fs(φ) =

f(φ)

1−F (φ∗s), if φ ≥ φ∗s,

0, otherwise

The value of a firm is v(φ) = max

0, φs(φ)ηs

. Thus, the free entry condition, (1− F (φ∗s)) π =

wηsfEs , can be written as follows:

∫ φ∗∗s

φ∗s

[(φPsφP∗s

)σCs −1

− 1

]fPs dF (φ)+

∫ ∞φ∗∗s

[(κs(φ)

(φPsφP∗s

)σCs −1

− 1

)fPs − fLs

]dF (φ) = ηsf

Es (22)

In equilibrium, the mass of successful entrants equals the mass of exiting firms: [1− F (φ∗s)]Ms =

ηsMs, where Ms is total number of firms and ME the constant mass of entering firms. Also,

free entry implies that total payments to labor used in entry must equal aggregate profits,

NE = MEfEs = Mπ = Π. Finally, R − Π = w(NP + NL) + rK +∑S

s=1 pXs Xs. Thus, labor

11

market clearing condition implies that N = NP + NL + NE = R − rK −∑S

s=1 pXs Xs, where we

set w = 1 as the numeraire.

The model has simple aggregation properties, as in Melitz (2003). The sectoral price index

can be written as:

Ps = M1

1−σCss ps(φs), (23)

where

ps(φs) =µCsφPs

(w

αNs

)αNs ( r

αKs

)αKs ( pXsαXs

)1−αNs −αKs,

φPs =

[MNL

s

Ms

(φP,NLs

)σCs −1

+ML

s

Ms

(φP,Ls

)σCs −1] 1

σCs −1

, (24)

φP,NLs =

[∫ φ∗∗s

φ∗s

(φP)σCs −1 dF (φ)

F (φ∗∗s )− F (φ∗s)

] 1

σCs −1

, (25)

φP,Ls =

[∫ ∞φ∗∗s

(φP (1 + τs(φ))

)σCs −1 dF (φ)

1− F (φ∗∗s )

] 1

σCs −1

, (26)

where MNLs = (1− ξLs )Ms and ML

s = ξLsMs is the mass of successful entry firms that do not select

and select into lobbying activity, respectively.

Given the first-order conditions of firms, we get that

Ks =αKsµCs

Rs

r,

Ns =αNsµCs

Rs

w, (27)

Xs =αXsµCs

Rs

pXs,

where Rs = Msrs is aggregate revenues of sector s.

E.2 Solution Algorithm of the Model

The steps taken to solve the model are the following:

1. Use the zero-profit condition from lobbying in (19) and the free entry condition in (22) to

recover φ∗s, φ∗∗s .

12

2. Compute ξLs , φP,Ls and φP,NLs from (21), (26) and (25), respectively.

3. Compute rs and πs from (20).

4. Compute Rs from (27) and Ms = Rs/rs.

5. Compute φPs from (24).

6. Compute Ps and P from (23) and (16), respectively.

7. Compute aggregate welfare, Y , from (1), and Ys from (17).

E.3 Proofs

This subsection presents the main proofs of the propositions in the paper.

Proof of Proposition 1. The first order condition of firms’ intensive margin lobbying decision is

the following:

δs(φLs ls(φ))δsrs(φ) = wls(φ)

By taking logs and rearranging one arrives to Equation (6).

Proof of Proposition 2. Firms’ first order conditions imply that the marginal revenue product of

factors are the following:

MRPNs(φ) ≡ ∂rs(φ)

∂ns(φ)=σCs − 1

σCsαNs

rs(φ)

ns(φ)=

w

1 + τs(φ)

MRPKs(φ) ≡ ∂rs(φ)

∂ks(φ)=σCs − 1

σCsαKs

rs(φ)

ks(φ)=

pK

1 + τs(φ)

MRPMs(φ) ≡ ∂rs(φ)

∂ms(φ)=σCs − 1

σCsαMs

rs(φ)

ms(φ)=

pXs1 + τs(φ)

Define aggregate labor used in variable costs at the sector level as Ns =∫ns(φ)Xsfs(dφ) and

similar objects for capital and intermediate inputs. Define also the average marginal revenue prod-

ucts of labor as MRPN s = 1∫MRPNs(φ)

rs(φ)PsYs

Msfs(dφ)and similar objects for capital and intermediate

13

inputs. Using these relationships, the standard monopolistic competition pricing and the standard

CES ideal price index, one has the following:

Ys = ΦPs N

αNss KαKs

s MαMss

ΦPs = X

1

σCs −1s

(NPs

Ns

)αNs (MRPN s

w

)αNs (MRPKs

pK

)αKs (MRPM s

pXs

)αMs [∫ (φP (1 + τs(φ))

)σCs −1fs(dφ)

] 1

σCs −1

Finally, define TFPR at the firm and sector level, respectively, as

TFPRs(φ) =(MRPNs(φ)

w

)αNs (MRPKs(φ)pK

)αKs (MRPMs(φ)pXs

)αMsand

TFPRs =(MRPNs

w

)αNs (MRPKs

pK

)αKs (MRPMs

pXs

)αMs, then one has the result:

ΦPs = X

1

σCs −1s

(NPs

Ns

)αNs [∫ (φP

TFPRs

TFPRs(φ))

)σCs −1

fs(dφ)

] 1

σCs −1

Proof of Proposition 3. Using the first order conditions of the problem stated in Equations (9)-(12)

and assuming that w = 1, one has the following:

a

P

[∂T

∂τ(φ)+

∂P

∂τ(φ)Y

]=(φL)σL−1

σL

(l(φ)

L

) 1

σL ∂l(φ)

∂τ(φ)fL(φ)

This highlights that in setting τ(φ), the government compares the benefit of obtaining more

lobbying expenditures and affecting the household’s welfare. The latter is a combination of affect-

ing the household’s income through changes in T and the price index P . Using the constraints in

Equations (9)-(12) and rearranging, one arrives to the result of Equation (13)

Appendix F Reduced Form Analysis

In this appendix we present robustness evidence to the IV strategy implemented in Section 4. We

present two types of robustness. The first, varies the timing of the weights used in the instrument

to weight the relevance of committees for firms. In the benchmark, we used the committee weights

14

that are lagged one period before we committee membership changes. We repeat the benchmark

result in the top panel of Table F.1. The middle and bottom panel of this table uses weights

lagged two and three years, respectively. One can see that the results are largely robust to this

variation. The second, uses value added as the outcome, rather than sales and profits. Table F.2

shows the main table using weights in t− 1, t− 2 and t− 3. The positive correlation in the OLS

and causal effect in the IV also holds when using value added. Furthermore, the direction of the

bias works in the same way as with sales.

15

Log Sales Log Profits

(1) (2) (3) (4) (5) (6) (7) (8)

Log Lobby 0.0436∗∗∗ 0.0500∗∗∗ 0.319∗∗∗ 0.227∗∗∗ 0.0241∗∗ 0.0468∗∗∗ 0.322∗∗∗ 0.236∗

(0.00891) (0.0117) (0.0754) (0.0795) (0.00948) (0.0130) (0.114) (0.124)

Firm and Year FE X X X X X X X X

State-Year FE X X X X X X X X

Model OLS OLS IV IV OLS OLS IV IV

F-Stat 21.30 20.20 14.30 13.10

Sample All Post 2007 All Post 2007 All Post 2007 All Post 2007

Weight Lag t-1 t-1 t-1 t-1

N 15332 8796 15332 8796 10369 6064 10369 6064

Log Sales Log Profits

(1) (2) (3) (4) (5) (6) (7) (8)

Log Lobby 0.0423∗∗∗ 0.0500∗∗∗ 0.378∗∗∗ 0.304∗∗∗ 0.0244∗∗ 0.0468∗∗∗ 0.488∗∗∗ 0.371∗∗∗

(0.00895) (0.0117) (0.0877) (0.0872) (0.00977) (0.0130) (0.147) (0.128)

Firm and Year FE X X X X X X X X

State-Year FE X X X X X X X X

Model OLS OLS IV IV OLS OLS IV IV

F-Stat 15.20 18.70 9.300 9.900

Sample All Post 2007 All Post 2007 All Post 2007 All Post 2007

Weight Lag t-2 t-2 t-2 t-2

N 14542 8796 14542 8796 9885 6064 9885 6064

Log Sales Log Profits

(1) (2) (3) (4) (5) (6) (7) (8)

Log Lobby 0.0448∗∗∗ 0.0500∗∗∗ 0.393∗∗∗ 0.381∗∗∗ 0.0303∗∗∗ 0.0468∗∗∗ 0.610∗∗∗ 0.560∗∗

(0.00962) (0.0117) (0.103) (0.126) (0.00977) (0.0130) (0.216) (0.234)

Firm and Year FE X X X X X X X X

State-Year FE X X X X X X X X

Model OLS OLS IV IV OLS OLS IV IV

F-Stat 11.40 16.40 6 6.900

Sample All Post 2007 All Post 2007 All Post 2007 All Post 2007

Weight Lag t-3 t-3 t-3 t-3

N 13710 8796 13710 8796 9344 6064 9344 6064

Table F.1: Different Baseline Weights for IV: This table presents the OLS and IV betweenlobbying expenditures and firms’ sales and profits. It shows robustness to the main table, usingdifferent committee weights. Besides presenting the baseline regression, it presents robustnessusing weights from t−2 and t−3. Profits are defined as sales minus wage bills, capital expendituresand intermediate input expenditures. All regressions have firm, year and state-year fixed effects.Standard errors are double clustered at firm and year level. *** p<0.01, ** p<0.05, * p<0.1

16

Log Sales Log VA

(1) (2) (3) (4) (5) (6) (7) (8)

Log Lobby 0.0436∗∗∗ 0.0500∗∗∗ 0.319∗∗∗ 0.227∗∗∗ 0.0251∗∗∗ 0.0299∗∗ 0.287∗∗∗ 0.177∗∗

(0.00891) (0.0117) (0.0754) (0.0795) (0.00667) (0.00948) (0.0916) (0.0686)

Firm and Year FE X X X X X X X X

State-Year FE X X X X X X X X

Model OLS OLS IV IV OLS OLS IV IV

F-Stat 21.30 20.20 14.50 17.10

Sample All Post 2007 All Post 2007 All Post 2007 All Post 2007

Weight Lag t-1 t-1 t-1 t-1

N 15332 8796 15332 8796 10851 6097 10851 6097

Log Sales Log VA

(1) (2) (3) (4) (5) (6) (7) (8)

Log Lobby 0.0423∗∗∗ 0.0500∗∗∗ 0.378∗∗∗ 0.304∗∗∗ 0.0255∗∗∗ 0.0299∗∗ 0.392∗∗∗ 0.271∗∗∗

(0.00895) (0.0117) (0.0877) (0.0872) (0.00684) (0.00948) (0.123) (0.0925)

Firm and Year FE X X X X X X X X

State-Year FE X X X X X X X X

Model OLS OLS IV IV OLS OLS IV IV

F-Stat 15.20 18.70 8.400 7.900

Sample All Post 2007 All Post 2007 All Post 2007 All Post 2007

Weight Lag t-2 t-2 t-2 t-2

N 14542 8796 14542 8796 10299 6097 10299 6097

Log Sales Log VA

(1) (2) (3) (4) (5) (6) (7) (8)

Log Lobby 0.0448∗∗∗ 0.0500∗∗∗ 0.393∗∗∗ 0.381∗∗∗ 0.0282∗∗∗ 0.0299∗∗ 0.564∗∗∗ 0.514∗∗∗

(0.00962) (0.0117) (0.103) (0.126) (0.00694) (0.00948) (0.195) (0.188)

Firm and Year FE X X X X X X X X

State-Year FE X X X X X X X X

Model OLS OLS IV IV OLS OLS IV IV

F-Stat 11.40 16.40 6.700 9.200

Sample All Post 2007 All Post 2007 All Post 2007 All Post 2007

Weight Lag t-3 t-3 t-3 t-3

N 13710 8796 13710 8796 9697 6097 9697 6097

Table F.2: Different Baseline Weights for IV: This table presents the OLS and IV betweenlobbying expenditures and firms’ sales and value-added. It shows robustness to the main table,using different committee weights. Besides presenting the baseline regression, it presents robust-ness using weights from t− 2 and t− 3. Value added is defined as sales minus intermediate inputexpenditures. All regressions have firm, year and state-year fixed effects. Standard errors aredouble clustered at firm and year level. *** p<0.01, ** p<0.05, * p<0.1

17